chore(shipped-layer): drop review-round narrative from the installer comments - #1359
Merged
Merged
Conversation
…comments Sibling sweep of the S1b round-3 MINOR 2 finding (#1356), which removed one such retelling from `_py_json_rules`. That fix named a class, and the class had nine members, not one — a wider grep over `setup.d/**` and `packages/core/templates/**` surfaced eight more sites where our internal review history is narrated inside files that install into consumer projects. Consumers read these seams to understand their own install. «MAJOR B (W-7)», «S2 rework round 1», «R2», «round-1 rework» and a path into `.superpowers/sdd/` are our process artefacts: the wave numbers index nothing a consumer can open, and the round numbers describe a review conversation they were never part of. Every substantive rationale is preserved verbatim — only the round/wave/severity markers and the one dead internal report path are removed: - `10-skills.sh` — env+ contour surface; the design-SSOT citation that carries the actual reason (spec:211 «pipeline presets, status, …») stays. - `10-skills.sh` — the §1j relocation note; «consolidate to ONE ship-point» stays as the reason, and the call-chain argument below it is untouched. - `85-worktree-scripts.sh` — getff-work.sh ship-alongside rationale (spec A9) stays. - `lib.sh` — the concrete leak example stays (it is the useful part); «(round-1 rework)» becomes plain observation. - `46-cargo.sh` / `47-go.sh` — «§3a option B / §6 fork 2» design pointers stay; the «MAJOR B (W-7)» prefix goes. - `45-python.sh` ×3 — probe-proven silent-unenforcement finding stays (the dead `.superpowers/sdd/s2-task-2-report.md` pointer goes with the round marker); the fragment-per-rule and manifest-path rationales stay, the latter restated so it explains the CURRENT design rather than narrating the defect that preceded it. Design pointers into internal kickoff/spec sections (`§3a`, `§6 fork 2`, `spec A8/A9`) are a NEIGHBOURING class — also unreachable for a consumer, but they carry design intent rather than review history. Left alone deliberately; removing them would rewrite meaning, not strip narrative. Verified on host: install snapshot 15/15 byte-identical (no delivered bytes move — these are installer-layer comments with no shipped twin), `bash -n` clean on all six edited files, local CI sweep green, make self-audit green. Prior-art: skipped — comment-only sweep removing review narrative from shipped files, no behaviour change, no new capability and no new dependency.
artyhoo
added a commit
that referenced
this pull request
Aug 10, 2026
…ually do (#1363) `packages/core/hooks/getff-work.test.ts` was flaky/red on the vitest 5000ms default: every case spawns a real `bash scripts/getff-work.sh`, which does `git worktree add` + node_modules wiring + a package-manager install inside the new worktree. On clean staging c1848e8 it ran 4 failed / 3 passed, all four `Test timed out in 5000ms`. In CI it was marginal — CC-DEFERRAL failed on PR #1359's first run and went green on re-run, having already blocked one unrelated comment-only PR. Timing the wrapper directly on a cold worktree measured 30.8s total (dominant leg: the step-2 `npm ci` inside the new worktree), so the 30_000 SLOW_SHELL_MS value the sibling suites use is itself too tight here. getff-work gets 60_000 — ~2x the measured cold cost, with precedent in principles/20-bundle-classification .paired-negative.test.ts and the root vitest.config.ts, and well under that file's documented 120_000 ceiling. Its inner spawnSync guard is named SPAWN_GUARD_MS so the "test timeout must outlive the spawn guard" inequality is visible. A full-suite sweep (`vitest run hooks/ skills/`) showed the same defect class across the siblings: 70 failed, 69 of them `timed out in 5000ms`. Fixed at 30_000 (the existing SLOW_SHELL_MS convention): hooks/create-worktree.test.ts, hooks/worktree-setup.test.ts, hooks/dup-detect-empty-arg.test.ts, hooks/priority-score-branch-matcher.test.ts, hooks/priority-score-reconstruct-stub.test.ts, skills/dup-detect.test.ts, skills/dup-detect-skip-closed.test.ts plus two CI-gated files that carried only PARTIAL per-test coverage — the untimed cases (C1/C2 tier-resolution; the fake-SHA paired-negative) were the ones going red, and `test:hooks` does not go green under load without them: hooks/validate-prompt.test.ts, hooks/check-worker-dispatch-channel.test.ts Verification (clean tree, macOS, 2026-08-10): - getff-work.test.ts alone: 7 passed (7), slowest 32.5s - `vitest run hooks/ skills/`: hooks/ failures 15 -> 0; total 70 -> 38 - `tsc --noEmit`: clean Timeout configuration only — no assertion, test-body, or `it()` line changed; `git diff` contains zero touched expect/toBe/toMatch/toContain lines. Known residue, deliberately NOT in this diff (no drive-by per CLAUDE.md «PR strategy»): the 38 remaining timeouts live in four skills/ files outside the audited set — planner-discovery (12), plan-currency-check (11), -digest (6), -kickoff (5), run-helper (4). They are the same defect class, but skills/ is gated by no workflow (CI runs test:hooks / test:principles / ir / backends / composition / live-generation / render), so they block no PR. Prior-art: skipped — test timeout configuration only, no new capability, no new dependency and no new module; the values reuse the SLOW_SHELL_MS convention that already exists in this package. Co-authored-by: Test <test@example.com>
10 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Sibling sweep of the S1b round-3 MINOR 2 finding closed in #1356. That fix removed one audit-round retelling from
_py_json_rules; the finding named a class, and the class turned out to have nine members, not one. This PR removes the remaining eight from files that install into consumer projects.Consumers read these seams to understand their own install.
MAJOR B (W-7),S2 rework round 1,R2,round-1 reworkand a path into.superpowers/sdd/are our process artefacts: the wave numbers index nothing a consumer can open, and the round numbers describe a review conversation they were never part of.The sweep that found them is wider than the grep used when the observation was first surfaced (that one reported four sites; it missed the
MAJOR A/B (W-n)family entirely).Changes
Every substantive rationale is preserved verbatim — only round/wave/severity markers and one dead internal report path are removed.
10-skills.sh:113beta-delivery-ux S2 / R3,S5 kickoff §2 binding #3,at S2 rework round 1beta-program-design.md:211)10-skills.sh:325beta-delivery-ux S2 rework round 1 / R785-worktree-scripts.sh:46R2 (S2 rework round 1)lib.sh:82(round-1 rework)46-cargo.sh:252,47-go.sh:219MAJOR B (W-7) /§3a option B / §6 fork 2design pointers45-python.sh:245(S2-T2 review; .superpowers/sdd/s2-task-2-report.md §Fix round 1)45-python.sh:648MAJOR B (W-7):45-python.sh:670MAJOR A (W-8):Deliberately out of scope
Design pointers into internal kickoff/spec sections (
§3a,§6 fork 2,spec A8/A9) are a neighbouring class — also unreachable for a consumer, but they carry design intent rather than review history. Removing them would rewrite meaning, not strip narrative, so they are left alone.One
round-2 auditretelling remains at45-python.sh:520. That is the exact line #1356 fixes; this branch is cut from staging, so the fix is not visible here. Not duplicated on purpose — it lands with that PR.No gate backs this
There is no mechanical check that keeps review narrative out of shipped files, so this class will recur (
attention-is-not-a-mechanism.md §1— a convention whose only enforcement is «someone notices in review» is not a mechanism). Not built here: that would be a capability commit and a separate concern from the cleanup. Flagged for the maintainer's call.Prior-art consult
Prior-art:trailer. Not a capability commit: no new dependency, no new file, nothing underpackages/— the diff is comment text in sixsetup.d/files.Test plan
SNAPSHOT_MODE=compare bash tests/install-sh/snapshot.sh— 15 pass / 0 fail, byte-identical (no delivered bytes move; these are installer-layer comments with no shipped twin)bash -nclean on all six edited filesbash scripts/run-local-ci-sweep.sh— selectedshellcheckfor this diff, PASSmake self-audit— 28/28, 274 passed + 4 skipped, composition 46/46, principles 353 passed + 1 skipped, always-on 31297B within the 54000B ceilingsetup.d/**+packages/core/templates/**, except the45-python.sh:520line fix(getff-freshness-widening): S1b — close the two round-3 MINORs (emitted proof corpus · shipped-layer narrative) #1356 ownsNot re-run here: the full
npm test --workspace=@rules-as-tests/core --runsuite, which is red on this host independently of any change (documented in #1356). This diff is comment-only and moves no delivered byte, so the install snapshot is the load-bearing check.CI went red once on a flake — recorded rather than quietly re-run
The first CI run failed
Principles as meta-tests (Phase 2)onhooks/getff-work.test.ts > CC-DEFERRAL,Test timed out in 5000ms. That test name sits uncomfortably close to this diff —setup.d/85-worktree-scripts.shis the file that shipsgetff-work.sh— so it was checked rather than assumed:git diff -U0filtered to non-comment lines returns empty), andscripts/getff-work.shitself is not touched by this PR at all.c1848e88fein a fresh provisioned worktree with none of these commits: 4 failed / 3 passed, including the sameCC-DEFERRAL— all four failing on timeout (9550ms, 7800ms, 6966ms, 6245ms against a 5000ms limit), not on an assertion.So the gate was re-run on evidence of non-causation, not on hope. The underlying issue — a spawn-heavy test wired to a 5000ms timeout, red locally and marginal in CI — is pre-existing and out of scope here.
Provenance
n/a
Review findings
n/a
Fidelity verdict
FIDELITY: skipped — non-stage follow-up sweep of a class named by a MINOR finding in #1356; comment-only, no kickoff substrate governs it and no behaviour changes.
Parked questions
n/a
§1.7 Self-discipline check (REQUIRED if PR touches discipline-bearing files)
§1.7 Skipped: mechanical maintenance — comment text removed from installer files; no discipline rule, principle, or skill is introduced or extended, and no changed path matches the discipline trigger.