diff --git a/.gitignore b/.gitignore
index 630c23b7..d1f24c9f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -481,6 +481,13 @@ FodyWeavers.xsd
 # 10's Native AOT publish-and-run proof) - regenerated by that script, never committed.
 .local-nuget-feed-nunit-aot-smoke/
 
+# Raw per-run outputs from skills/compono-workspace/ eval iterations - transient, regenerated by
+# each run. grading.json/timing.json/benchmark.json/feedback.json ARE committed (this repo's own
+# established evaluation-evidence convention, see skills/compono-workspace/benchmarks/), only the
+# bulky raw outputs/ directories are excluded.
+skills/compono-workspace/iteration-*/**/outputs/
+skills/compono-workspace/skill-snapshot/
+
 # macOS
 .DS_Store
 .DS_Store?
diff --git a/README.md b/README.md
index e0b17ae2..5263a9aa 100644
--- a/README.md
+++ b/README.md
@@ -128,8 +128,11 @@ instead of opening a public issue.
 
 ## Status
 
-Compono is under active development. APIs are experimental until the
-first public preview.
+Compono is publicly released, with stable `0.x` packages on NuGet. It's
+still pre-`1.0`, so the public API can still change before the `1.0`
+compatibility boundary — see
+[Installation](https://layeredcraft.github.io/compono/getting-started/installation/)
+for exact package versions.
 
 ## License
 
diff --git a/docs/getting-started/installation.md b/docs/getting-started/installation.md
index 3c4855a8..2c9fe125 100644
--- a/docs/getting-started/installation.md
+++ b/docs/getting-started/installation.md
@@ -7,14 +7,14 @@ composed test method parameters/theories. That's two packages — `Compono`
 plus whichever test-framework integration matches your test host:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.XunitV3 --prerelease
+dotnet add package Compono
+dotnet add package Compono.XunitV3
 # or, for TUnit:
-dotnet add package Compono.TUnit --prerelease
+dotnet add package Compono.TUnit
 # or, for MSTest (MSTest.TestFramework 4.0.0+ - see the Compono.MSTest Package Guide):
-dotnet add package Compono.MSTest --prerelease
+dotnet add package Compono.MSTest
 # or, for NUnit (NUnit 3.14.0+ - see the Compono.NUnit Package Guide):
-dotnet add package Compono.NUnit --prerelease
+dotnet add package Compono.NUnit
 ```
 
 This tutorial's assertions (`.Should()`, throughout this site's own
@@ -29,17 +29,15 @@ dotnet add package AwesomeAssertions
 Add the rest of the ecosystem as your tests need it:
 
 ```bash
-dotnet add package Compono.NSubstitute --prerelease         # automatic substitute composition
-dotnet add package Compono.Bogus --prerelease               # semantic fake data (names, emails, ...)
-dotnet add package Compono.DependencyInjection --prerelease # row.AsServiceProvider() bridge
-dotnet add package Compono.Http --prerelease                # TestHttpHandler for HttpClient tests
+dotnet add package Compono.NSubstitute         # automatic substitute composition
+dotnet add package Compono.Bogus               # semantic fake data (names, emails, ...)
+dotnet add package Compono.DependencyInjection # row.AsServiceProvider() bridge
+dotnet add package Compono.Http                # TestHttpHandler for HttpClient tests
 ```
 
-Every package targets `net8.0`/`net9.0`/`net10.0`/`net11.0`. Until the first stable `1.0`
-release, every published version is a `0.x.y-preview.N` prerelease — the
-`--prerelease` flag (or an explicit `<PackageReference Version="0.x.y-preview.N" />`)
-is required for `dotnet add package`/NuGet restore to pick it up at all,
-since a plain `dotnet add package` skips prerelease versions by default. See
+Every package targets `net8.0`/`net9.0`/`net10.0`/`net11.0` and has a
+stable release, so a plain `dotnet add package` picks up the right version
+for all of them with no extra flag. See
 [Package Guides](../packages/index.md) for what each package is for and
 when to add it.
 
diff --git a/docs/migrating-from-autofixture.md b/docs/migrating-from-autofixture.md
index dbf34b36..352c2dce 100644
--- a/docs/migrating-from-autofixture.md
+++ b/docs/migrating-from-autofixture.md
@@ -67,17 +67,16 @@ AutoFixture users migrating xUnit tests need:
   anonymous values.
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.XunitV3 --prerelease
-dotnet add package Compono.NSubstitute --prerelease
-dotnet add package Compono.Bogus --prerelease
+dotnet add package Compono
+dotnet add package Compono.XunitV3
+dotnet add package Compono.NSubstitute
+dotnet add package Compono.Bogus
 ```
 
 Install matching versions of every Compono package you add — mixing
 versions across packages isn't supported; see
 [Package Guides: Version Compatibility](packages/index.md#version-compatibility).
-See [Installation](getting-started/installation.md) for the full setup,
-including why `--prerelease` is required during public preview.
+See [Installation](getting-started/installation.md) for the full setup.
 
 ## Quick concept map
 
diff --git a/docs/packages/compono-bogus.md b/docs/packages/compono-bogus.md
index 4f1352c3..9bbd3894 100644
--- a/docs/packages/compono-bogus.md
+++ b/docs/packages/compono-bogus.md
@@ -12,8 +12,8 @@ You want composed string/data members to look like plausible real values
 placeholder strings:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.Bogus --prerelease
+dotnet add package Compono
+dotnet add package Compono.Bogus
 ```
 
 If anonymous placeholder values are fine for your tests, you don't need
diff --git a/docs/packages/compono-dependencyinjection.md b/docs/packages/compono-dependencyinjection.md
index 7aaab599..af49b232 100644
--- a/docs/packages/compono-dependencyinjection.md
+++ b/docs/packages/compono-dependencyinjection.md
@@ -19,8 +19,8 @@ to enumerate and manually register every dependency a system under test
 might ask for:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.DependencyInjection --prerelease
+dotnet add package Compono
+dotnet add package Compono.DependencyInjection
 ```
 
 This package is deliberately **not** framework-specific. It doesn't
diff --git a/docs/packages/compono-http.md b/docs/packages/compono-http.md
index 7ca645d5..51b4fa3b 100644
--- a/docs/packages/compono-http.md
+++ b/docs/packages/compono-http.md
@@ -18,8 +18,8 @@ pipeline against a configured HTTP response — not substitute an
 application-level interface:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.Http --prerelease
+dotnet add package Compono
+dotnet add package Compono.Http
 ```
 
 If the seam under test is already an ordinary application interface
diff --git a/docs/packages/compono-logging.md b/docs/packages/compono-logging.md
index a9cacdc2..7491a405 100644
--- a/docs/packages/compono-logging.md
+++ b/docs/packages/compono-logging.md
@@ -19,8 +19,8 @@ message, structured properties, exception, scope — not substitute an
 unrelated application interface:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.Logging --prerelease
+dotnet add package Compono
+dotnet add package Compono.Logging
 ```
 
 This package depends only on `Compono` and
diff --git a/docs/packages/compono-mstest.md b/docs/packages/compono-mstest.md
index 387d8637..40ed8717 100644
--- a/docs/packages/compono-mstest.md
+++ b/docs/packages/compono-mstest.md
@@ -9,8 +9,8 @@ that composes test method parameters directly, instead of hand-building
 You write MSTest tests and want method parameters composed automatically:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.MSTest --prerelease
+dotnet add package Compono
+dotnet add package Compono.MSTest
 ```
 
 `Compono.MSTest` doesn't add an MSTest test host for you — it integrates
diff --git a/docs/packages/compono-nsubstitute.md b/docs/packages/compono-nsubstitute.md
index c97f05a7..22ff556f 100644
--- a/docs/packages/compono-nsubstitute.md
+++ b/docs/packages/compono-nsubstitute.md
@@ -12,8 +12,8 @@ abstract classes) that you'd otherwise create with `Substitute.For<T>()` by
 hand:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.NSubstitute --prerelease
+dotnet add package Compono
+dotnet add package Compono.NSubstitute
 ```
 
 If none of your composed types have interface/delegate/abstract-class
diff --git a/docs/packages/compono-nunit.md b/docs/packages/compono-nunit.md
index 86170c8e..4ff8943a 100644
--- a/docs/packages/compono-nunit.md
+++ b/docs/packages/compono-nunit.md
@@ -9,8 +9,8 @@ attribute that composes test method parameters directly, without needing
 You write NUnit tests and want method parameters composed automatically:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.NUnit --prerelease
+dotnet add package Compono
+dotnet add package Compono.NUnit
 ```
 
 `Compono.NUnit` doesn't add an NUnit test host for you — it integrates with
diff --git a/docs/packages/compono-testdoubles.md b/docs/packages/compono-testdoubles.md
index 2a65abe8..f44d96e6 100644
--- a/docs/packages/compono-testdoubles.md
+++ b/docs/packages/compono-testdoubles.md
@@ -11,8 +11,8 @@ generated double, without pulling in `Compono.NSubstitute`'s runtime proxy
 dependency (or when you need the composed path to survive `PublishAot`):
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.TestDoubles --prerelease
+dotnet add package Compono
+dotnet add package Compono.TestDoubles
 ```
 
 `Compono.TestDoubles` is not a general-purpose mocking framework — see
diff --git a/docs/packages/compono-tunit.md b/docs/packages/compono-tunit.md
index 9438ebd4..03e82553 100644
--- a/docs/packages/compono-tunit.md
+++ b/docs/packages/compono-tunit.md
@@ -9,8 +9,8 @@ custom data source generator.
 You write TUnit tests and want method parameters composed automatically:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.TUnit --prerelease
+dotnet add package Compono
+dotnet add package Compono.TUnit
 ```
 
 `Compono.TUnit` doesn't add a TUnit test host for you — it integrates with
diff --git a/docs/packages/compono-xunitv3.md b/docs/packages/compono-xunitv3.md
index 8cc93d03..3e6fdebc 100644
--- a/docs/packages/compono-xunitv3.md
+++ b/docs/packages/compono-xunitv3.md
@@ -10,8 +10,8 @@ You write xUnit v3 tests (`xunit.v3` + the Microsoft Testing Platform
 runner) and want theory parameters composed automatically:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.XunitV3 --prerelease
+dotnet add package Compono
+dotnet add package Compono.XunitV3
 ```
 
 `Compono.XunitV3` doesn't add an xUnit v3 test host for you — it integrates
diff --git a/docs/packages/compono.md b/docs/packages/compono.md
index 87448e2b..1a427e4f 100644
--- a/docs/packages/compono.md
+++ b/docs/packages/compono.md
@@ -15,7 +15,7 @@ dependency of `Compono.XunitV3`/`Compono.NSubstitute`/`Compono.Bogus`, so
 directly.
 
 ```bash
-dotnet add package Compono --prerelease
+dotnet add package Compono
 ```
 
 ## What it gives you
diff --git a/docs/packages/index.md b/docs/packages/index.md
index db77dade..ee014124 100644
--- a/docs/packages/index.md
+++ b/docs/packages/index.md
@@ -18,10 +18,9 @@ need the first two.
 | [`Compono.Http`](compono-http.md) | `TestHttpHandler` — a reflection-free `HttpMessageHandler` test double: `OnGet`/`OnPost`/etc. + `When(...)` matching, strict unmatched-request behavior, registration-handle verification. | Your test needs to exercise the real `HttpClient` pipeline against a configured HTTP response, instead of substituting an application-level interface. |
 | [`Compono.Logging`](compono-logging.md) | `UseLogging()` — `ILogger`/`ILogger<T>` compose as a hand-written `CapturingLogger`/`CapturingLogger<T>`, with structured-property extraction, real scope tracking, and `Verify()` verification. Generation is on by default once installed. | Your composed type takes an `ILogger`/`ILogger<T>` dependency and the test wants to assert what was logged. |
 
-Every package targets `net8.0`/`net9.0`/`net10.0`/`net11.0` and, until the
-first stable `1.0` release, publishes as a `0.x.y-preview.N` prerelease —
-see [Installation](../getting-started/installation.md) for the exact
-`dotnet add package` commands and why `--prerelease` is required. See
+Every package targets `net8.0`/`net9.0`/`net10.0`/`net11.0` and has a
+stable release — see [Installation](../getting-started/installation.md)
+for the exact `dotnet add package` commands. See
 [ADR-0038](../adr/0038-net8-net9-explicit-multi-target.md) for why `net8.0`/
 `net9.0` were added alongside the existing `net10.0`/`net11.0` window.
 
@@ -38,14 +37,14 @@ integration matches your test host — `Compono.XunitV3` for xUnit v3,
 for NUnit:
 
 ```bash
-dotnet add package Compono --prerelease
-dotnet add package Compono.XunitV3 --prerelease
+dotnet add package Compono
+dotnet add package Compono.XunitV3
 # or, for TUnit:
-dotnet add package Compono.TUnit --prerelease
+dotnet add package Compono.TUnit
 # or, for MSTest:
-dotnet add package Compono.MSTest --prerelease
+dotnet add package Compono.MSTest
 # or, for NUnit:
-dotnet add package Compono.NUnit --prerelease
+dotnet add package Compono.NUnit
 ```
 
 Add `Compono.NSubstitute` and/or `Compono.Bogus` independently, as your
diff --git a/docs/research/0022-1-0-readiness-adversarial-review.md b/docs/research/0022-1-0-readiness-adversarial-review.md
new file mode 100644
index 00000000..406ef2f8
--- /dev/null
+++ b/docs/research/0022-1-0-readiness-adversarial-review.md
@@ -0,0 +1,240 @@
+# [RESEARCH-0022] 1.0 Readiness Adversarial Review
+
+**Status:** Complete. No product/API-design blocker found. Five
+release-mechanics/hardening items accepted as required pre-1.0 work
+(tracked below, resolved and verified across two follow-up rounds); five
+further items explicitly considered and rejected as unnecessary gates.
+
+## Why this exists
+
+Final release-readiness gate for Compono 1.0, run as an adversarial
+review whose job was to actively try to prove Compono should **not** ship
+1.0 yet — not to summarize prior plans' conclusions. Covered: public API
+freeze, deferred pre/post-1.0 decisions (disposal, async composition,
+framework-binder duplication), source-generation/runtime contracts,
+Native AOT/trimming, package correctness (all 11 publishable packages),
+framework integrations, canonical samples, documentation sync, CI/package
+validation gates, release/versioning readiness, and dogfooding.
+
+## Method
+
+Direct repository investigation plus four parallel research forks
+(public-API/generator-contract audit; package-correctness/release
+audit; framework-integration/canonical-sample audit; AOT/CI-gate audit),
+cross-checked against a fresh clean build, full test run, and two real
+dogfood-consumer runs (a live external consumer, `trivia-platform`,
+restored against freshly-packed current-`HEAD` Compono packages via
+`scripts/dogfood-validate.sh`). Three of the four forks were lost
+mid-run to a session rate limit; their load-bearing findings were
+recovered by direct inspection rather than re-run (see the original
+review's own "checks not executed" section for the exact gaps).
+
+## Original findings (as reported, unmodified)
+
+**No BLOCKER found under the review's own criteria** (a finding blocks
+1.0 only if fixing it after 1.0 likely requires a breaking change, is a
+correctness defect in a supported scenario, is an unstable
+generated-code/runtime contract, misstates a package's compatibility
+contract, or is a genuine trimming/AOT gap where Compono claims
+compatibility). One process-integrity finding was flagged at
+blocker-equivalent severity because it undermines confidence in every
+other gate the review relied on:
+
+- **[F0] Required status checks on `main` covered only `build / build`**
+  (blocker-equivalent, not a numbered Release Task) — `Package
+  Validation` and `AOT Validation` (the gates PLAN-0061/0062 built
+  specifically for the 1.0 boundary) were not required checks, so either
+  could fail red and a PR could still merge.
+
+Six items were originally classified as RELEASE TASKS, numbered F1-F6
+below (this numbering is this record's own stable identifier scheme, not
+part of the original review's prose — introduced to remove the ambiguity
+a reviewer flagged in an earlier draft that referred back to items only
+as "Item N above"):
+
+1. **[F1]** `breaking-change` label auto-resolves to `v1.0.0` with no
+   separate decision point (ADR-0031 Amendment 5, already deliberately
+   wired).
+2. **[F2]** Packed-consumer smoke-test coverage in
+   `package-validation.yaml` covers only 5 of 11 packages
+   (XunitV3/TUnit/MSTest/NUnit/TestDoubles); `Compono`, `Http`, `Logging`,
+   `DependencyInjection`, `NSubstitute`, `Bogus` rely on static nuspec
+   inspection + AOT-smoke packing only.
+3. **[F3]** Install docs universally instruct `--prerelease`.
+4. **[F4]** ADR-0031's "rolling two-TFM window" prose appeared to
+   disagree with the actual four-TFM (`net8.0;net9.0;net10.0;net11.0`)
+   build matrix.
+5. **[F5]** `scripts/dogfood-validate.sh` has an undocumented Docker
+   prerequisite (its default consumer's repository tests use
+   Testcontainers).
+6. **[F6]** 20 xUnit-analyzer warnings (`xUnit1031`/`xUnit1051`) in
+   `Compono.DependencyInjection.Tests` aren't gated as errors.
+
+Deferred decisions re-verified safely post-1.0, unchanged from prior
+research: disposal/lifetime ownership ([RESEARCH-0015](0015-disposal-ownership-research.md),
+Outcome C), async composition ([RESEARCH-0016](0016-async-composition-viability-research.md),
+Outcome C), framework-binder duplication ([RESEARCH-0019](0019-framework-binder-duplication-spike-scope.md),
+still scoped-not-executed, no second drift bug found on independent
+re-check of the negative-seed guard across all four binders).
+
+## Disposition after product-owner review
+
+The product owner accepted the original review's core conclusion (no
+product/API-design blocker) without reopening it, then narrowed the six
+Release Task items:
+
+**Accepted as required pre-1.0 work (4):**
+
+- **[F0]** (blocker-equivalent) — required checks. **Resolved and
+  verified live**: `LayeredCraft/.github` PR #5 (merged) opts Compono out
+  of safe-settings' `required_status_checks.contexts` management via the
+  same `{{EXTERNALLY_DEFINED}}` marker `dynamodb-efcore-provider`/
+  `devops-templates` already use — confirmed via
+  `gh api repos/LayeredCraft/compono/branches/main/protection`, before
+  (`["build / build"]` only, unaffected by the merge) and after applying
+  a narrow `PATCH .../branches/main/protection/required_status_checks`
+  (not a full protection-object replace) setting
+  `["build / build", "package-validation", "aot-gate"]` with
+  `strict: true`. Re-read after the change: exactly those three contexts
+  present, no literal `{{EXTERNALLY_DEFINED}}` check registered, and every
+  other branch-protection field (reviews, `enforce_admins`, linear
+  history, force-push/delete, conversation resolution, signatures)
+  byte-identical to before.
+- **[F3]** — stale `--prerelease` install guidance. Resolved across three
+  passes:
+  1. First pass: audited and corrected `docs/getting-started/installation.md`,
+     all 11 `docs/packages/*.md` guides, `docs/packages/index.md`,
+     `docs/migrating-from-autofixture.md`, and `docs/troubleshooting/faq.md`.
+     Correction: **most packages already have a stable `0.9.0` release**
+     on nuget.org (verified live) — `--prerelease` was already unnecessary
+     and wrong for `Compono`/`Compono.XunitV3`/`Compono.TUnit`/
+     `Compono.NSubstitute`/`Compono.Bogus`/`Compono.DependencyInjection`/
+     `Compono.Http`/`Compono.Logging`/`Compono.TestDoubles` *today*, not
+     merely "once 1.0 ships" as the original review assumed. This pass
+     kept `Compono.MSTest`/`Compono.NUnit` on `--prerelease`, since
+     neither had a stable release (verified live against nuget.org).
+  2. A Codex review of PR #131 correctly flagged a bug this introduced:
+     `Compono.MSTest`/`Compono.NUnit` each pin their `Compono` dependency
+     to their own **exact** package version (`[%(ProjectVersion)]` bracket
+     syntax in each `.csproj`'s `PinProjectReferenceVersionsExact`
+     target) — verified directly against the packed `.nuspec` on
+     nuget.org: `Compono.MSTest`/`Compono.NUnit` `0.10.0-preview.101` both
+     declare `<dependency id="Compono" version="[0.10.0-preview.101]" />`.
+     The docs' shared "`dotnet add package Compono`" line (resolving
+     stable `0.9.0`) followed by `Compono.MSTest`/`Compono.NUnit
+     --prerelease` (resolving `0.10.0-preview.101`) taught an invalid
+     pairing — NuGet cannot satisfy an exact `[0.10.0-preview.101]`
+     dependency with an already-resolved `0.9.0` core.
+  3. Before that fix shipped, the product owner directed a different,
+     simpler resolution: `Compono.MSTest`/`Compono.NUnit` are about to go
+     stable together with every other package immediately after this PR
+     merges, so the docs now install all six packages
+     (`Compono`/`XunitV3`/`TUnit`/`MSTest`/`NUnit` plus the four
+     independent add-ons) the same way, with no `--prerelease` anywhere
+     and no stable/preview distinction — correct once that release lands,
+     and this record notes plainly that the docs are written slightly
+     ahead of the actual publish rather than pretending otherwise.
+- **[F4]** — the ADR-0031 TFM discrepancy. **Investigated and found to be a
+  false positive**, not a real drift: ADR-0031's own **Amendment 3**
+  (2026-08-10) already records, in the same file, that ADR-0038
+  superseded the original two-TFM framing and that
+  `net8.0;net9.0;net10.0;net11.0` are all actively-tracked TFMs. The
+  original review's finding came from reading only the ADR's original
+  "Decision Outcome" prose (2026-08-04) without reading its own later
+  amendments in the same document. No ADR/documentation change was
+  needed — the decision record was already accurate. Nothing in
+  current-facing docs (README, package guides) states a TFM count that
+  would need correcting either.
+- **[F5]** — dogfood Docker prerequisite. Resolved: `scripts/dogfood-validate.sh`'s
+  usage text now states the prerequisite and its failure mode explicitly.
+- The README's "Status" section ("APIs are experimental until the first
+  public preview") was also already stale *today* — Compono has stable
+  `0.9.0` packages, not merely "preview." Corrected to state plainly:
+  publicly released, stable `0.x` packages exist, still pre-`1.0` so the
+  API can still change.
+- **New in this round: Compono skill eval/workspace structure** — brought
+  into alignment with the current [Agent Skills eval-workspace
+  convention](https://agentskills.io/skill-creation/evaluating-skills).
+  Authored `evals.json` moved from the sibling `skills/compono-evals/`
+  directory (a 2026-08-07 fix, at the time correct for the ecosystem
+  convention as it existed then — see PLAN-0035's own "Real defect"
+  note) into `skills/compono/evals/evals.json`, matching the doc's
+  canonical `<skill>/evals/evals.json` layout; the generated-workspace
+  half (`benchmarks/`, all historical, preserved as-is) moved to a new
+  sibling `skills/compono-workspace/`, matching the doc's
+  `<skill>-workspace/` sibling convention, with a new
+  `skills/compono-workspace/README.md` documenting the current
+  `iteration-N/`/`with_skill`+`old_skill`/`timing.json`/`grading.json`/
+  `benchmark.json`/`feedback.json` shape for future runs. **Trade-off
+  reintroduced, flagged rather than silently accepted:** this move puts
+  `evals.json` (51KB, 46 scenarios) back inside the directory `npx skills
+  add` copies verbatim to every consumer installing the `compono` skill —
+  exactly the bloat PLAN-0035's 2026-08-07 move was built to avoid. The
+  current agentskills.io specification and eval-workflow doc were both
+  checked directly and neither documents any installer-side exclusion
+  mechanism for `evals/`; the doc's own canonical example ships
+  `evals/evals.json` (and any `evals/files/`) as part of the installed
+  skill. This is a genuine, upstream-level tension between "evals belong
+  in the skill directory" (current convention) and "evals shouldn't ship
+  to every consumer" (this repo's original, still-valid concern) — not
+  something this pass invented or silently resolved. Implemented as
+  directed because the instruction was explicit and precisely specified,
+  not exploratory; flagged here as a decision the product owner may want
+  to revisit if consumer-side skill-install size becomes a real
+  complaint (e.g. by asking upstream for a `.skillsignore`-equivalent, or
+  reconsidering the trade-off).
+
+**Explicitly rejected as unnecessary pre-1.0 gates (5):**
+
+- **[F2]** (additional packed-consumer smoke projects for the remaining 6
+  packages) — the product owner determined the original Logging finding
+  was specifically a ProjectReference-vs-packed-MSBuild-assets
+  development concern (already correctly scoped and handled — see
+  [PLAN-0061](../plans/0061-pre-1-0-cleanup-and-consolidation.md) Phase
+  2 and the packed asset's own inline documentation), and that the
+  packed nupkg's assets for those 6 packages are already inspected by
+  `inspect-packed-nupkgs.sh`. No new concrete evidence of an actual
+  uncovered packaging failure mode was presented, so six more
+  packed-consumer scenarios were not added.
+- **[F6]** (20 xUnit-analyzer warnings) — no shipped-product correctness
+  impact; left for separate follow-up.
+- Additional release-drafter machinery around `breaking-change → major`
+  — that mapping (ADR-0031 Amendment 5) is intentional, already-decided
+  SemVer behavior for leaving `0.x`; publishing itself stays
+  human-controlled regardless. No new gate added.
+- Re-litigating disposal/lifetime ownership, async composition, or
+  framework-binder consolidation — the adversarial review found no
+  reason for any of these to block 1.0; not reopened.
+- Additional AOT projects without a contradicted support claim (e.g. for
+  `Compono.Bogus`/`Compono.DependencyInjection`, which document no AOT
+  claim either way) — not added; nothing regresses by adding this later.
+
+## Final disposition
+
+**READY FOR 1.0**, pending only the release-execution act itself. The
+product owner has stated intent to cut `1.0.0` for every package
+immediately once PR #131 merges — F3's final documentation state (§
+above) is written for that near-term stable world, not the transient
+state that existed while `Compono.MSTest`/`Compono.NUnit` were still
+preview-only. All five accepted pre-1.0/hardening items above are
+resolved and verified, including the required-status-check change now
+confirmed live on `main`. No item originally listed as a Release Task,
+and later rejected, should be read as a mandatory gate — each rejection
+above is a deliberate product-owner decision, not an oversight.
+
+## Links
+
+- Original adversarial review — conducted in-session; no separate
+  research record existed for it prior to this one, which supersedes it
+  as the authoritative write-up of both the original findings and their
+  disposition.
+- [PLAN-0061](../plans/0061-pre-1-0-cleanup-and-consolidation.md),
+  [PLAN-0062](../plans/0062-package-validation-gap-fixes.md) — the
+  pre-1.0 work this review verified.
+- [ADR-0031](../adr/0031-public-preview-release-and-versioning-policy.md) —
+  release/versioning policy, including Amendment 5 (`0.x` graduation)
+  referenced above.
+- [ADR-0038](../adr/0038-net8-net9-explicit-multi-target.md) — the
+  four-TFM decision ADR-0031 Amendment 3 already reconciles against.
+- `LayeredCraft/.github` PR #5 — safe-settings repo-managed
+  required-checks opt-out for `compono`.
diff --git a/docs/troubleshooting/faq.md b/docs/troubleshooting/faq.md
index 35d20a76..dcfb5837 100644
--- a/docs/troubleshooting/faq.md
+++ b/docs/troubleshooting/faq.md
@@ -56,12 +56,17 @@ instance every call) rather than a fresh one.
 
 ## Is there a stable `1.0` release yet?
 
-Not yet — every published version during public preview is a
-`0.x.y-preview.N` prerelease (`--prerelease` required to install). See
-[MVP Non-goals](../mvp.md#mvp-non-goals) and
+Not yet, but every package already has a stable release — a plain
+`dotnet add package` picks it up with no extra flag. See
+[Installation](../getting-started/installation.md) for exact commands,
+[MVP Non-goals](../mvp.md#mvp-non-goals), and
 [ADR-0031](../adr/0031-public-preview-release-and-versioning-policy.md)
-for the compatibility policy this implies: a breaking change bumps the
-minor version (`0.X+1.0`), not a major version, until `1.0`.
+for the compatibility policy this implies. Note ADR-0031's own Amendment 5:
+a breaking change now bumps the major version (the ordinary SemVer
+`breaking-change → major` mapping), not the minor version — the ADR's
+original "bumps the minor version until `1.0`" override was a deliberate,
+temporary `0.x`-era guard that's since been lifted now that Compono is
+ready to leave the `0.x` line.
 
 ## Next
 
diff --git a/scripts/dogfood-validate.sh b/scripts/dogfood-validate.sh
index 49ef61b6..f5a9ace3 100755
--- a/scripts/dogfood-validate.sh
+++ b/scripts/dogfood-validate.sh
@@ -85,6 +85,13 @@ Exit code is 0 only if packing, restore, version verification (every requested p
 the exact freshly-packed version - not a stale cache hit, and not a mix of freshly-packed and
 previously published versions), and the consumer's full test suite all succeed. The consumer
 repo's git working tree is left exactly as it was found, regardless of outcome.
+
+Prerequisite: the default consumer (trivia-platform) includes Testcontainers-backed repository
+tests that require a running Docker daemon. Without one, those tests fail with
+DockerUnavailableException and this script exits non-zero even when every Compono package under
+test is fine - that's an environment gap, not a Compono regression. Start Docker (or point
+--consumer-repo/DOGFOOD_CONSUMER_REPO at a consumer that doesn't need it) before relying on this
+script's exit code as a pass/fail signal.
 EOF
 }
 
diff --git a/skills/compono-workspace/README.md b/skills/compono-workspace/README.md
new file mode 100644
index 00000000..f4d9d81e
--- /dev/null
+++ b/skills/compono-workspace/README.md
@@ -0,0 +1,66 @@
+# `compono` skill eval workspace
+
+Generated eval-execution evidence for the `compono` agent skill
+(`skills/compono/`), kept alongside the skill directory per the
+[Agent Skills eval-workspace convention](https://agentskills.io/skill-creation/evaluating-skills) —
+authored eval definitions live in `skills/compono/evals/evals.json`;
+everything here is generated output, never authored skill content.
+
+## Structure
+
+```
+compono-workspace/
+├── skill-snapshot/          # cp -r of skills/compono/ taken before an edit, when
+│                             # comparing an iteration against the previous skill
+│                             # revision rather than against no skill at all
+├── iteration-N/
+│   ├── eval-<case-name>/
+│   │   ├── with_skill/
+│   │   │   ├── outputs/     # files the run produced
+│   │   │   ├── timing.json  # { "total_tokens": ..., "duration_ms": ... }
+│   │   │   └── grading.json # assertion_results + summary, per evals.json's
+│   │   │                     # "assertions"/"expectations" for that case
+│   │   └── old_skill/       # or without_skill/ - see "Baseline choice" below
+│   │       ├── outputs/
+│   │       ├── timing.json
+│   │       └── grading.json
+│   ├── benchmark.json       # aggregated pass-rate/time/token stats for the iteration
+│   └── feedback.json        # { "<eval-case-name>": "<human review note>", ... }
+└── benchmarks/               # pre-convention historical runs (2026-08-07 through
+                               # 2026-09-03) - kept as-is; each is its own dated
+                               # README/benchmark.md, not restructured into the
+                               # iteration-N/ shape above
+```
+
+## Baseline choice
+
+`compono` is an established skill under active iteration, not a new skill
+being evaluated from zero — per the convention above, a revision compares
+against a `skill-snapshot/` of the previous version (`old_skill/`), not
+against no skill (`without_skill/`), unless the run is specifically
+measuring whether having the skill at all is worth it (in which case use
+`without_skill/` and say so explicitly in that iteration's directory).
+
+## Clean-context execution
+
+Each with-skill/baseline run in an iteration starts from a fresh
+subagent (the `Agent` tool's `general-purpose` type, or a fresh Claude
+Code session if run outside this repo) — never a continuation of the
+skill-development conversation. This is what makes a run's output
+attributable to `SKILL.md` itself rather than to prior conversational
+state. Each run is given: the skill path (or none, or the snapshot path),
+the eval's `prompt`, its `files` (if any), and a distinct output directory
+so concurrent with-skill/baseline runs can never collide.
+
+## Grading
+
+Grade each case's `assertions`/`expectations` (from `evals/evals.json`)
+against that run's `outputs/`, with concrete evidence per PASS/FAIL —
+per this repo's own established practice (see `benchmarks/`), grading is
+currently done by direct reasoning against the skill's actual reference
+content, not an automated content-grading harness;
+`.agents/skills/skill-creator/scripts/run_eval.py` is a *trigger*-eval
+runner (does the skill activate at all), not a content grader. Prefer a
+deterministic script wherever an assertion is mechanically checkable
+(valid syntax, a specific API name present/absent); reserve reasoning-based
+grading for assertions that genuinely require it.
diff --git a/skills/compono-evals/benchmarks/2026-08-07/README.md b/skills/compono-workspace/benchmarks/2026-08-07/README.md
similarity index 96%
rename from skills/compono-evals/benchmarks/2026-08-07/README.md
rename to skills/compono-workspace/benchmarks/2026-08-07/README.md
index 6426b3d9..31a1147a 100644
--- a/skills/compono-evals/benchmarks/2026-08-07/README.md
+++ b/skills/compono-workspace/benchmarks/2026-08-07/README.md
@@ -1,7 +1,7 @@
 # Benchmark run — 2026-08-07
 
 A `/skill-creator`-*style* eval run against all 18 scenarios in
-`../../evals.json` (superseding the 6-scenario manual spot-check recorded
+`../../../compono/evals/evals.json` (superseding the 6-scenario manual spot-check recorded
 earlier in PLAN-0035) — not `/skill-creator`'s full documented workflow;
 see "Known limitations" below for exactly where it's lighter-weight.
 With-skill and baseline (`without_skill`, no access to the skill's
@@ -89,7 +89,7 @@ follow-up for whoever next revises `evals.json`, per this repo's
 
 ## Superseded eval manifest
 
-`../../evals.json` now contains 26 scenarios. This benchmark remains an
+`../../../compono/evals/evals.json` now contains 26 scenarios. This benchmark remains an
 accurate historical record for scenarios 1-18 only; it does **not** measure
 the later additions for parameterized profiles, NUnit, TUnit, generated
 test doubles, dependency injection, custom providers, `CreateMany`, or
diff --git a/skills/compono-evals/benchmarks/2026-08-07/benchmark.json b/skills/compono-workspace/benchmarks/2026-08-07/benchmark.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/benchmark.json
rename to skills/compono-workspace/benchmarks/2026-08-07/benchmark.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/benchmark.md b/skills/compono-workspace/benchmarks/2026-08-07/benchmark.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/benchmark.md
rename to skills/compono-workspace/benchmarks/2026-08-07/benchmark.md
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-1-use-compono-to-create-the-request-model-/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-10-my-integration-test-spins-up-a-webapplic/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-11-why-does-my-test-throw-nullreferenceexce/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-12-i-want-to-register-iclock-twice-in-my-pr/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-13-should-i-mark-my-entire-domain-model-wit/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-14-write-a-generic-fibonacci-function-in-c-/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-15-this-project-already-uses-compono-and-co/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-16-compono-can-t-compose-httpclient-in-my-t/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-17-compono-keeps-throwing-compositionexcept/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-18-give-me-a-realistic-looking-customer-wit/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-2-why-is-compono-failing-to-compose-this-t/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-3-convert-this-xunit-test-that-manually-bu/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-4-i-want-the-same-nsubstitute-dependency-r/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-5-use-bogus-for-the-email-address-but-let-/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-6-create-a-theory-using-compono-with-a-fix/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-7-why-did-adding-composable-not-fix-this-i/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-8-set-up-a-test-project-from-scratch-using/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/with_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/with_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/with_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/with_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/without_skill.json b/skills/compono-workspace/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/without_skill.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/without_skill.json
rename to skills/compono-workspace/benchmarks/2026-08-07/grading/eval-9-this-nsubstitute-test-doubles-a-service-/without_skill.json
diff --git a/skills/compono-evals/benchmarks/2026-08-25/README.md b/skills/compono-workspace/benchmarks/2026-08-25/README.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-25/README.md
rename to skills/compono-workspace/benchmarks/2026-08-25/README.md
diff --git a/skills/compono-evals/benchmarks/2026-08-25/benchmark.md b/skills/compono-workspace/benchmarks/2026-08-25/benchmark.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-25/benchmark.md
rename to skills/compono-workspace/benchmarks/2026-08-25/benchmark.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/benchmark.json b/skills/compono-workspace/benchmarks/2026-08-27/benchmark.json
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/benchmark.json
rename to skills/compono-workspace/benchmarks/2026-08-27/benchmark.json
diff --git a/skills/compono-evals/benchmarks/2026-08-27/benchmark.md b/skills/compono-workspace/benchmarks/2026-08-27/benchmark.md
similarity index 98%
rename from skills/compono-evals/benchmarks/2026-08-27/benchmark.md
rename to skills/compono-workspace/benchmarks/2026-08-27/benchmark.md
index 4608ee10..16b87c1e 100644
--- a/skills/compono-evals/benchmarks/2026-08-27/benchmark.md
+++ b/skills/compono-workspace/benchmarks/2026-08-27/benchmark.md
@@ -4,7 +4,7 @@ Focused regression run for the AWS Secrets Manager Provider TestDoubles matching
 
 ## Scope
 
-New evals added in `skills/compono-evals/evals.json`:
+New evals added in `skills/compono/evals/evals.json`:
 
 - 28 — `IConfigurationBuilder` argument-filtered verification
 - 29 — `IAmazonSecretsManager` argument-matched configuration
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/eval-28-with-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-28-with-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/eval-28-with-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-28-with-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/eval-29-with-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-29-with-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/eval-29-with-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-29-with-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/eval-30-with-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-30-with-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/eval-30-with-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-30-with-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/eval-31-with-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-31-with-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/eval-31-with-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/eval-31-with-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-28-old-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-28-old-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-28-old-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-28-old-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-29-old-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-29-old-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-29-old-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-29-old-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-30-old-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-30-old-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-30-old-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-30-old-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-31-old-skill.md b/skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-31-old-skill.md
similarity index 100%
rename from skills/compono-evals/benchmarks/2026-08-27/outputs/old_skill/eval-31-old-skill.md
rename to skills/compono-workspace/benchmarks/2026-08-27/outputs/old_skill/eval-31-old-skill.md
diff --git a/skills/compono-evals/benchmarks/2026-08-28/README.md b/skills/compono-workspace/benchmarks/2026-08-28/README.md
similarity index 96%
rename from skills/compono-evals/benchmarks/2026-08-28/README.md
rename to skills/compono-workspace/benchmarks/2026-08-28/README.md
index 77a08db8..e6a60c6f 100644
--- a/skills/compono-evals/benchmarks/2026-08-28/README.md
+++ b/skills/compono-workspace/benchmarks/2026-08-28/README.md
@@ -1,7 +1,7 @@
 # Benchmark run — 2026-08-28
 
 Scoped, single-eval run for the new `Share<T>()` eval added by PLAN-0056
-Task 6 (id 34 in `../../evals.json`) — not a full re-run of all 34
+Task 6 (id 34 in `../../../compono/evals/evals.json`) — not a full re-run of all 34
 scenarios. With-skill and without-skill (`without_skill`, told not to
 read `skills/compono/`) subagents ran independently against eval 34's
 prompt, graded manually against its `expectations` (5 assertions).
diff --git a/skills/compono-evals/benchmarks/2026-09-02/README.md b/skills/compono-workspace/benchmarks/2026-09-02/README.md
similarity index 97%
rename from skills/compono-evals/benchmarks/2026-09-02/README.md
rename to skills/compono-workspace/benchmarks/2026-09-02/README.md
index 1beeda52..e11b0565 100644
--- a/skills/compono-evals/benchmarks/2026-09-02/README.md
+++ b/skills/compono-workspace/benchmarks/2026-09-02/README.md
@@ -1,7 +1,7 @@
 # Benchmark run — 2026-09-02
 
 Scoped, two-eval run for the new `Compono.MSTest` evals added by
-PLAN-0057 task group 14 (ids 40-41 in `../../evals.json`) — not a full
+PLAN-0057 task group 14 (ids 40-41 in `../../../compono/evals/evals.json`) — not a full
 re-run of all 41 scenarios. Graded by the implementing session directly
 (same limitation as the 2026-08-28 run), reasoning through a with-skill
 answer (grounded in `skills/compono/references/mstest.md`'s actual
diff --git a/skills/compono-evals/benchmarks/2026-09-03/README.md b/skills/compono-workspace/benchmarks/2026-09-03/README.md
similarity index 98%
rename from skills/compono-evals/benchmarks/2026-09-03/README.md
rename to skills/compono-workspace/benchmarks/2026-09-03/README.md
index 9bf77210..93a86cf5 100644
--- a/skills/compono-evals/benchmarks/2026-09-03/README.md
+++ b/skills/compono-workspace/benchmarks/2026-09-03/README.md
@@ -1,7 +1,7 @@
 # Benchmark run — 2026-09-03
 
 Scoped, five-eval run for the new `Compono.NUnit` evals added by
-PLAN-0059 (ids 42-46 in `../../evals.json`) — not a full re-run of all 46
+PLAN-0059 (ids 42-46 in `../../../compono/evals/evals.json`) — not a full re-run of all 46
 scenarios. Graded by the implementing session directly (same limitation
 as the 2026-08-28/2026-09-02 runs — no automated content-grading harness
 exists for this repo's `evals.json` prompt/expected_output format;
diff --git a/skills/compono-workspace/iteration-6/benchmark.json b/skills/compono-workspace/iteration-6/benchmark.json
new file mode 100644
index 00000000..bdde64cc
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/benchmark.json
@@ -0,0 +1,20 @@
+{
+  "purpose": "Structural validation of the skills/compono/evals/ + skills/compono-workspace/ relocation (RESEARCH-0022 follow-up) - proves eval discovery, clean-context execution, output separation, timing capture, and grading still work after the move. Single eval (id 1), not a full 46-scenario re-run; not intended to drive a SKILL.md change.",
+  "run_summary": {
+    "with_skill": {
+      "pass_rate": { "mean": 1.0 },
+      "tokens": { "mean": 53722 },
+      "duration_ms": { "mean": 54234 }
+    },
+    "without_skill": {
+      "pass_rate": { "mean": 0.667 },
+      "tokens": { "mean": 31624 },
+      "duration_ms": { "mean": 23619 }
+    },
+    "delta": {
+      "pass_rate": 0.333,
+      "tokens": 22098,
+      "duration_ms": 30615
+    }
+  }
+}
diff --git a/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/grading.json b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/grading.json
new file mode 100644
index 00000000..38c28649
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/grading.json
@@ -0,0 +1,25 @@
+{
+  "assertion_results": [
+    {
+      "text": "Uses only real Compono APIs (Composer.Create(...) to build the composer, composer.Create<T>()/CreateMany<T>() as instance methods, [Compose], etc.), nothing invented, and distinguishes programmatic composition from a Compose row",
+      "passed": true,
+      "evidence": "Option A uses [Theory][Compose] as a real Compono.XunitV3 attribute; Option B uses `Composer.Create()` (static factory) then the instance method `Composer.Create<CreateOrderRequest>()` - no invented API, and the two options are explicitly labeled 'inside an xUnit v3 theory' vs 'inside the test body (arrange step, not a theory parameter)', a clear programmatic-vs-Compose-row distinction."
+    },
+    {
+      "text": "Does not add [Composable] unless there is a genuine discovery-gap reason",
+      "passed": true,
+      "evidence": "Explicit 'Why not other approaches' section: 'No [Composable] on CreateOrderRequest... the type is composed directly at a visible call site... adding the attribute would be speculative, not a fix for anything broken.'"
+    },
+    {
+      "text": "Does not introduce Activator.CreateInstance or any reflection-based fallback",
+      "passed": true,
+      "evidence": "Explicit: 'No Activator.CreateInstance or other reflection... Compono is source-generated by design (ADR-0001) and has no reflection fallback.'"
+    }
+  ],
+  "summary": {
+    "passed": 3,
+    "failed": 0,
+    "total": 3,
+    "pass_rate": 1.0
+  }
+}
diff --git a/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/timing.json b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/timing.json
new file mode 100644
index 00000000..28117b47
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/with_skill/timing.json
@@ -0,0 +1,4 @@
+{
+  "total_tokens": 53722,
+  "duration_ms": 54234
+}
diff --git a/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/grading.json b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/grading.json
new file mode 100644
index 00000000..5a6b3392
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/grading.json
@@ -0,0 +1,28 @@
+{
+  "assertion_results": [
+    {
+      "text": "Uses only real Compono APIs (Composer.Create(...) to build the composer, composer.Create<T>()/CreateMany<T>() as instance methods, [Compose], etc.), nothing invented, and distinguishes programmatic composition from a Compose row",
+      "passed": false,
+      "evidence": "Invents `[ComposableTest]` (not a real Compono.XunitV3 attribute - the real attribute is `[Compose]` on a `[Theory]`) and `new Composer()` (Compono's Composer is constructed via the static `Composer.Create()` factory, not a public constructor)."
+    },
+    {
+      "text": "Does not add [Composable] unless there is a genuine discovery-gap reason",
+      "passed": true,
+      "evidence": "Never mentions or adds [Composable] - a baseline pass by omission rather than by correct reasoning about the attribute's purpose."
+    },
+    {
+      "text": "Does not introduce Activator.CreateInstance or any reflection-based fallback",
+      "passed": true,
+      "evidence": "No reflection introduced - a baseline pass by omission rather than by correct reasoning about Compono's source-generation design."
+    }
+  ],
+  "summary": {
+    "passed": 2,
+    "failed": 1,
+    "total": 3,
+    "pass_rate": 0.667
+  },
+  "additional_observations": [
+    "Recommends post-hoc mutation ('request.Quantity = 5;' after Create<T>()) as the way to pin specific values - directly contradicts the with-skill answer's explicit member-rule/CompositionBuilder guidance and the general Compono philosophy against composing then mutating. Not covered by an eval-1 assertion, but a real quality gap flagged as human-review feedback (see feedback.json)."
+  ]
+}
diff --git a/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/timing.json b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/timing.json
new file mode 100644
index 00000000..6d462128
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/eval-1-use-compono-to-create-the-request-model/without_skill/timing.json
@@ -0,0 +1,4 @@
+{
+  "total_tokens": 31624,
+  "duration_ms": 23619
+}
diff --git a/skills/compono-workspace/iteration-6/feedback.json b/skills/compono-workspace/iteration-6/feedback.json
new file mode 100644
index 00000000..0ca2871f
--- /dev/null
+++ b/skills/compono-workspace/iteration-6/feedback.json
@@ -0,0 +1,3 @@
+{
+  "eval-1-use-compono-to-create-the-request-model": "With-skill answer is correct and matches the eval's assertions. No SKILL.md change proposed from this single-eval structural-validation run - not a full iteration."
+}
diff --git a/skills/compono-evals/evals.json b/skills/compono/evals/evals.json
similarity index 100%
rename from skills/compono-evals/evals.json
rename to skills/compono/evals/evals.json