Statusline 2.1.0 — one cost ledger, ten modules, width-aware rendering

.claude-plugin/marketplace.json

@@ -6,7 +6,7 @@
   },
   "metadata": {
     "description": "Session optimizer, split into three independently installable plugins: context-guard (Stop-hook context budget + memory-writer checkpoint subagent + subagent spend tracker), refine-gate (UserPromptSubmit prompt-binding gate + /refine skill), and statusline (multi-line status bar + install skill).",
-    "version": "2.0.0"
+    "version": "2.1.0"
   },
   "plugins": [
     {
@@ -57,8 +57,8 @@
     {
       "name": "statusline",
       "source": "./plugins/statusline",
-      "description": "Multi-line Claude Code statusline with RGB-gradient context bars tied to per-model checkpoint thresholds (shared with context-guard), monthly cost tracking, per-session telemetry, rate-limit gauges, and live subagent spend. Ships an install skill that copies the bundled assets into ~/.claude and wires settings.json.",
-      "version": "2.0.0",
+      "description": "Multi-line status bar: a discrete heat-track context bar tied to per-model checkpoint thresholds (shared with context-guard), one deduplicated cost ledger that prices each session from its own transcript plus its subagents, 5h/7d rate-limit gauges with burn-rate pacing (the projection of usage at reset, not just the percentage), per-session telemetry (tok/s, compactions, prompt-cache countdown), and terminal-width fitting that drops the lowest-priority segments instead of letting the host truncate. Words, not emoji. Ships an install skill.",
+      "version": "2.1.0",
       "author": {
         "name": "Clement Deust",
         "email": "admin@ai-architect.tools"

.github/workflows/ci.yml

@@ -19,20 +19,39 @@ jobs:
         run: |
           python -m pip install --upgrade pip
           pip install pytest
-          sudo apt-get update
-          sudo apt-get install -y shellcheck
+
+      # Pinned rather than taken from apt: the distro build is several releases
+      # behind, and the two versions do not report the same findings — it raises
+      # SC2317 where 0.11.0 raises SC2329, and SC2015 on `A && B || true` which
+      # 0.11.0 no longer flags. An unpinned linter makes a local run greener
+      # than the gate, which is how three findings reached CI unseen.
+      - name: Install shellcheck 0.11.0
+        run: |
+          v=v0.11.0
+          curl -fsSL "https://github.com/koalaman/shellcheck/releases/download/${v}/shellcheck-${v}.linux.x86_64.tar.xz" \
+            | sudo tar -xJf - --strip-components=1 -C /usr/local/bin "shellcheck-${v}/shellcheck"
+          shellcheck --version
 
       - name: Refine gate contract tests
         run: pytest tests/test_refine_gate.py -v
 
       - name: Subagent tracker tests
         run: pytest tests/test_subagent_usage.py -v
 
-      - name: Statusline unit tests
+      - name: Statusline heat-track tests
         run: bash tests/statusline/test_heat_rgb.sh
 
-      - name: Shellcheck statusline script
-        run: shellcheck plugins/statusline/assets/statusline-command.sh
+      - name: Statusline fitting, pace and module tests
+        run: bash tests/statusline/test_fit_and_pace.sh
+
+      - name: Shellcheck statusline renderer, modules, ledger and tests
+        run: |
+          shellcheck plugins/statusline/assets/statusline-command.sh
+          shellcheck plugins/statusline/assets/statusline-lib/*.sh
+          shellcheck plugins/statusline/assets/costs.sh
+          # The suites are shell too, and a checker that skips them lets the
+          # code that guards the renderer rot unwatched.
+          shellcheck tests/statusline/*.sh
 
       - name: Validate marketplace, plugin, and hook JSON
         run: |
@@ -45,3 +64,4 @@ jobs:
           done
           python -m json.tool plugins/statusline/assets/statusline-budget.json > /dev/null
           python -m json.tool plugins/statusline/assets/ctxguard-thresholds.json > /dev/null
+          python -m json.tool plugins/statusline/assets/pricing.json > /dev/null

CHANGELOG.md

@@ -5,6 +5,83 @@ All notable changes to this project are documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.1.0] - 2026-07-26
+
+Statusline only. No change to context-guard or refine-gate.
+
+### BREAKING
+
+- **`statusline-costs.py` is removed.** Every dollar figure now comes from one
+  ledger, `costs.sh` over `~/.claude/statusline-costs.jsonl`. The Python
+  aggregator summed every assistant line of every transcript and over-counted
+  ~2.2x, because Claude Code re-logs one API response 2-3 times (streaming /
+  tool continuation); a correct total must deduplicate on
+  `message.id:requestId` before pricing. Measured over 172 local transcripts:
+  $3438.84 deduped vs $7645.04 raw. `costs.sh` also prices the full recursive
+  `<transcript>/subagents/` subtree, so Task, worktree-isolated and workflow
+  agents are all billed — none of them appear in Claude Code's own
+  `.cost.total_cost_usd`. Installs carrying a stale `~/.claude/statusline-costs.py`
+  should delete it; the SessionStart hook now does so.
+- **The renderer ships as a directory.** `statusline-command.sh` is a
+  composition root that sources `statusline-lib/*.sh`, which must be installed
+  next to it. A missing module is a hard failure that names the file rather
+  than a partial statusline. `$STATUSLINE_LIB` overrides the location.
+
+### Added
+
+- `costs.sh` + `pricing.json` as bundled assets (the ledger and its prices).
+- Terminal-width fitting: each line is held to 85% of the probed width, and
+  `fit_line` drops whole trailing segments — lowest priority first — rather
+  than letting the host truncate mid-word and cost the block a row. Width is
+  probed from the controlling tty, `$COLUMNS`, then `tput cols` (only when
+  stdout is a terminal); `$STATUSLINE_COLS` overrides.
+- Verbosity preset now falls back to `s` below 90 columns. The config's
+  `"size"` is a preference capped by width; `$STATUSLINE_SIZE` remains a hard
+  pin honoured at any width.
+- **Pace** on both rate-limit windows: used% over the share of the window
+  elapsed, which is the linear projection of usage at reset (`1.0x` lands
+  exactly on the cap). The percentage carries the worse of the absolute and
+  pace severities; the pace figure carries its own. Nothing is printed below
+  10% of the window elapsed, where the extrapolation is not informative.
+- Tests: `tests/statusline/test_fit_and_pace.sh` (46 tests — fitting, pace,
+  severity, the width probe, preset resolution, the module loader's failure
+  path, and the §4.1 size cap), `tests/statusline/measure_widths.sh`
+  (per-preset width measurement).
+- CI now shellchecks `tests/statusline/*.sh` as well as the assets. A checker
+  that skips the suites lets the code guarding the renderer rot unwatched.
+
+### Changed
+
+- The identity line renders `model | dir | effort | thinking`. Segment order is
+  priority order under width fitting, and the previous order put `dir` last,
+  which made the working directory the first thing dropped on a narrow
+  terminal.
+- The renderer is split into ten modules, one concern per file, none over 500
+  lines (`rules/coding-standards.md` §4.1; the single file had reached 1022).
+  Behaviour-preserving: verified byte-identical across a 34-configuration
+  preset x width golden render.
+- Docs: both READMEs described an emoji-based layout that the word-based
+  design-system rendering had already replaced.
+
+### Fixed
+
+- `stat -f` is BSD-only; every mtime read now goes through `file_mtime`, which
+  tries the BSD and GNU spellings. On Linux the bare call silently read 0,
+  forcing a cache refresh on every invocation.
+- A non-numeric `used_percentage` (`"n/a"`) rendered as a healthy 0%: awk reads
+  an unquoted non-numeric token as an uninitialised variable. Values are now
+  validated before awk sees them.
+- `tput cols` returns terminfo's blind 80 when stdout is a pipe, which is how
+  the host captures the renderer — it is now consulted only when stdout is a
+  terminal, so IDE and web sessions no longer silently downgrade.
+- A non-numeric `$STATUSLINE_COLS` was printed straight through, breaking every
+  arithmetic width comparison downstream. The override is an escape hatch, not
+  an exemption from `probe_cols`'s postcondition: an invalid value now falls
+  through to the probes.
+- The terminal-size probe leaked `Device not configured` on every refresh with
+  no controlling tty: the failing redirection is reported by the shell itself,
+  so the whole group is now redirected, not just the command.
+
 ## [2.0.0] - 2026-07-22
 
 ### BREAKING

README.md

@@ -15,7 +15,7 @@ ones you want; none requires the others.
 |---|---|---|
 | [**context-guard**](plugins/context-guard) | `/plugin install context-guard@session-optimizer-marketplace` | A `Stop` hook enforces a per-model context budget: at the WARN threshold it writes a mechanical checkpoint stub and delegates persistence to a budgeted `memory-writer` subagent as a reflection pause; at the hard cap it forces checkpoint → `/clear` → resume. A `SubagentStop` tracker surfaces true session spend (main thread + subagents). |
 | [**refine-gate**](plugins/refine-gate) | `/plugin install refine-gate@session-optimizer-marketplace` | A `UserPromptSubmit` hook + `/refine` skill that bind vague prompt references ("the SSE solution", "like before", "still broken") to concrete artifacts with evidence, then select an execution strategy from a research-backed table — before any code is touched. |
-| [**statusline**](plugins/statusline) | `/plugin install statusline@session-optimizer-marketplace` | A multi-line status bar: RGB-gradient context bar tied to per-model checkpoint thresholds, cost, telemetry (tok/s, compactions, cache countdown), rate-limit gauges, live subagent spend. Ships an install skill — after installing, ask Claude to "install the statusline" and it wires everything. |
+| [**statusline**](plugins/statusline) | `/plugin install statusline@session-optimizer-marketplace` | A multi-line status bar: discrete heat-track context bar tied to per-model checkpoint thresholds, one deduplicated cost ledger covering subagent spend, telemetry (tok/s, compactions, cache countdown), rate-limit gauges with burn-rate pacing, and terminal-width fitting. Ships an install skill — after installing, ask Claude to "install the statusline" and it wires everything. |
 
 ## Why
 

plugins/statusline/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "statusline",
-  "description": "Multi-line Claude Code statusline (Catppuccin Mocha) with RGB-gradient context bars tied to per-model checkpoint thresholds, monthly cost tracking, per-session telemetry (tok/s, compactions, prompt-cache countdown), rate-limit gauges, and live subagent spend. Ships an install skill that copies the bundled assets into ~/.claude and wires settings.json.",
-  "version": "2.0.0",
+  "description": "Multi-line Claude Code statusline with a discrete heat-track context bar tied to per-model checkpoint thresholds, one deduplicated cost ledger covering subagent spend, per-session telemetry (tok/s, compactions, prompt-cache countdown), rate-limit gauges with burn-rate pacing, and terminal-width fitting. Words, not emoji: every value is named. Ships an install skill that copies the bundled assets into ~/.claude and wires settings.json.",
+  "version": "2.1.0",
   "author": {
     "name": "Clement Deust",
     "email": "admin@ai-architect.tools"