Skip to content

docs: fix stale --prerelease install guidance, document dogfood Docker prerequisite - #131

Merged
ncipollina merged 3 commits into
mainfrom
docs/1-0-readiness-followup
Sep 4, 2026
Merged

docs: fix stale --prerelease install guidance, document dogfood Docker prerequisite#131
ncipollina merged 3 commits into
mainfrom
docs/1-0-readiness-followup

Conversation

@ncipollina

Copy link
Copy Markdown
Contributor

Summary

  • Most packages already have a stable 0.9.0 release (verified live against nuget.org) — corrected --prerelease guidance across install docs, package guides, migration guide, and FAQ. Compono.MSTest/Compono.NUnit correctly keep --prerelease (no stable release yet for either).
  • Documented scripts/dogfood-validate.sh's Docker prerequisite (its default consumer's Testcontainers-backed tests need Docker; without it the script fails for reasons unrelated to Compono).
  • Added RESEARCH-0022 recording the 1.0 adversarial readiness review's findings and the product owner's disposition (accepted vs. explicitly rejected release tasks).
  • Investigated the ADR-0031 TFM-policy question; found it already resolved by the ADR's own Amendment 3 — no doc change needed there.

Test plan

  • No source (src/, test/) files changed — full Release build (0 errors) and full test suite (3483/3483 passed) from earlier in this review remain valid.
  • Verified live nuget.org versions for all 11 packages before editing any doc.
  • CI (docs build, package validation) — should be unaffected, doc/script-only change.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GMQEjFrUVGYkctE3ECfRuA

…r prerequisite

Most Compono packages already have a stable 0.9.0 release on nuget.org -
--prerelease was already wrong (not just "wrong once 1.0 ships") for
Compono/XunitV3/TUnit/NSubstitute/Bogus/DependencyInjection/Http/Logging/
TestDoubles. Compono.MSTest and Compono.NUnit have no stable release yet
and correctly keep --prerelease everywhere. Also documents
dogfood-validate.sh's Docker prerequisite (its default consumer's
Testcontainers-backed repository tests need a Docker daemon; without one
the script exits non-zero for reasons unrelated to Compono).

Adds RESEARCH-0022 recording the 1.0 adversarial readiness review's
original findings and the product owner's disposition of each: what was
accepted as required pre-1.0 work vs. explicitly rejected as unnecessary.

Investigated but made no change for: ADR-0031's TFM-policy wording - its
own Amendment 3 (2026-08-10) already reconciles the four-TFM
(net8.0/9.0/10.0/11.0) reality; the earlier finding read only the
original 2026-08-04 Decision Outcome prose without reading the
amendment.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GMQEjFrUVGYkctE3ECfRuA
@github-actions github-actions Bot added the type: docs Documentation change label Sep 4, 2026

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b10920280a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread docs/getting-started/installation.md
Comment thread docs/research/0022-1-0-readiness-adversarial-review.md Outdated
ncipollina and others added 2 commits September 3, 2026 21:21
…to current convention

README's "Status" section said "experimental until the first public
preview" - already false today (stable 0.9.0 packages exist). Corrected
to state Compono is publicly released with stable 0.x packages, still
pre-1.0.

Moved the compono skill's authored evals from the sibling
skills/compono-evals/ directory into skills/compono/evals/evals.json,
matching the current Agent Skills eval convention
(agentskills.io/skill-creation/evaluating-skills) - the previous location
was itself a deliberate 2026-08-07 fix for the ecosystem convention as it
existed then (see PLAN-0035). The generated-workspace half (all
historical benchmark runs, preserved as-is) moved to a new sibling
skills/compono-workspace/, with a new README documenting the current
iteration-N/with_skill+old_skill/timing.json/grading.json/benchmark.json/
feedback.json shape. Fixed relative-path links in historical benchmark
READMEs that broke from the move; left their narrative text untouched.

Ran one real eval (id 1) through the new structure end-to-end -
with-skill and no-skill subagents, separate output dirs, timing capture,
assertion grading, iteration benchmark - to prove discovery/execution/
grading all still work post-move, not to tune the skill itself.

Trade-off flagged in RESEARCH-0022: this reintroduces evals.json into
the directory npx skills add ships to every consumer, which the original
2026-08-07 move was built to avoid. Neither the current agentskills.io
specification nor its eval-workflow doc documents an installer-side
exclusion mechanism for evals/ - a genuine open tension, not something
this change resolves.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GMQEjFrUVGYkctE3ECfRuA
…te RESEARCH-0022

Codex review on PR #131 correctly flagged that Compono.MSTest/Compono.NUnit
each pin their Compono dependency to an exact package version
(PinProjectReferenceVersionsExact's [%(ProjectVersion)] bracket syntax,
confirmed against the packed .nuspec on nuget.org: both declare
<dependency id="Compono" version="[0.10.0-preview.101]" />), so the
previous docs' shared "dotnet add package Compono" (stable 0.9.0) ahead of
"Compono.MSTest/Compono.NUnit --prerelease" (0.10.0-preview.101) taught an
install sequence NuGet cannot actually satisfy.

Per direction: since every package is going stable together immediately
after this PR merges, resolved by dropping the preview distinction
entirely rather than teaching a soon-obsolete preview-pairing rule -
docs/getting-started/installation.md, docs/packages/compono-mstest.md,
docs/packages/compono-nunit.md, docs/packages/index.md,
docs/migrating-from-autofixture.md, docs/troubleshooting/faq.md, and
README.md now install/describe all packages the same way, with no
--prerelease anywhere.

Also fixed Codex's second finding: RESEARCH-0022 referred to findings only
as "Item N above", ambiguous once the numbered Release Task list existed
alongside an unnumbered blocker-equivalent finding. Introduced stable
[F0]-[F6] identifiers and updated every reference; preserved the original
adversarial evidence, only added identifiers and updated F3's disposition
with this round's two-step resolution (the exact-pin bug, then the
final stable-for-everyone direction).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GMQEjFrUVGYkctE3ECfRuA
@ncipollina
ncipollina merged commit d9db7b5 into main Sep 4, 2026
11 checks passed
@ncipollina
ncipollina deleted the docs/1-0-readiness-followup branch September 4, 2026 01:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

type: docs Documentation change

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant