Skip to content

docs(ssot): withdraw TaskMaster install from SSOT #84 (one-click-installer S4) - #449

Merged
artyhoo merged 1 commit into
stagingfrom
worktree-oci-s4-ssot
Jun 10, 2026
Merged

docs(ssot): withdraw TaskMaster install from SSOT #84 (one-click-installer S4)#449
artyhoo merged 1 commit into
stagingfrom
worktree-oci-s4-ssot

Conversation

@artyhoo

@artyhoo artyhoo commented Jun 10, 2026

Copy link
Copy Markdown
Owner

What

Amends SSOT row #84 (docs/meta-factory/prior-art-evaluations.md:154) with the TaskMaster-install withdrawal note, verbatim from plan Task 6 (docs/superpowers/plans/2026-05-31-one-click-installer.md, «Task 6: SSOT annotation») per one-click-installer kickoff §2 S4. Also bumps the row's Last reviewed cell 2026-05-27 → 2026-06-10 per the register's field contract. Amendment of the existing row — no new numbered entry, no renumbering.

Verify-trace

  • git branch --show-currentworktree-oci-s4-ssot
  • grep -c '^| 84 ' docs/meta-factory/prior-art-evaluations.md1
  • grep -n "withdrawn 2026-05-31" docs/meta-factory/prior-art-evaluations.md → single hit, line 154 (row 84)
  • grep '^| 84 ' … | grep -o '2026-05-27' | wc -l3 (was 4 — only Last reviewed changed; First seen, probe date, and the 2026-05-27-install-sh-k1-extension.md citation filename untouched)
  • Row stays one physical line; pipe count unchanged (9 → 9)
  • npx vitest run packages/core/principles/08-prior-art-cited.test.ts → 1 file / 7 tests passed
  • npx markdownlint-cli2 docs/meta-factory/prior-art-evaluations.mdSummary: 0 error(s)
  • CI on this PR

tests verified locally, CI wiring lands in S5

Scope

1 file: docs/meta-factory/prior-art-evaluations.md (row #84 cell amendment + Last-reviewed bump). No capability commit — docs-only edit per the CLAUDE.md capability-commit definition (no new dependency, no new file under packages/); no Prior-art: trailer required.

Note (orchestrator): branch was recreated server-side via the Git Data API on top of staging HEAD 9c10c68 (git SSH/HTTPS transport blocked by the known flaky tunnel; documented fallback). Local pre-push hooks therefore did not run on push — compensated by the worker's local principle-08 + markdownlint runs above and by CI on this PR.

§1.7 Forward-check applied

  • SSOT register contract (docs/meta-factory/prior-art-evaluations.md:70): «Do not edit existing rows in place except for Last reviewed updates per §3 step 1; substantive verdict changes are new commits with git history audit» — verdict cell (ADOPT) unchanged; the dated note append is the maintainer-approved plan-Task-6 amendment, and this commit is its git-history audit. Last reviewed bump per field spec (prior-art-evaluations.md:28).
  • Principle 08 (packages/core/principles/08-prior-art-cited.test.ts): run post-edit → 7/7 PASS (no broken SSOT references introduced).
  • Doc-authority (.claude/rules/doc-authority-hierarchy.md §3): no header change; row-cell edit only.
  • Note-internal claims verified: prior-art-evaluations.md:143 confirms row fix(hooks): aggregate full assistant turn for end-of-turn reminder #73 IS TaskMaster analyze-complexity — the appended «SSOT fix(hooks): aggregate full assistant turn for end-of-turn reminder #73 (TM vocabulary) unaffected» reference is accurate.

§1.7 Backward-check applied

Sweep: grep -rni 'taskmaster|task-master' across README/INSTALL*/CLAUDE.md/docs/.claude/rules/packages/scripts/install.sh/setup*/agents (node_modules + SSOT excluded). Findings:

  • install.sh / setup / setup.d / scripts / .claude/rules / agents: 0 hits — the live install path is already TM-free (S1 refactor(install): S1 — framework-only install.sh (one-click-installer Stage 1) #442 removal held).
  • S4-docs scope (README, NOT fixed here — sibling S4-docs sub-wave): README.md:121 («Optionally installs companion tools (Superpowers, TaskMaster) via claude plugin install»), README.md:127 («Install TaskMaster? [y/N]» prompt), README.md:131 (COMPANIONS=superpowers,task-master example), README.md:136 («TaskMaster install (install.sh:411)» — line ref now dangling). These four still claim TM installs via install.sh — handled by the sibling S4-docs PR in this stage.
  • Out-of-scope observations (closed historical artifacts, correctly untouched): research-patches dated 2026-05-25…2026-05-30 (e.g. 2026-05-27-install-sh-k1-extension.md:81, 2026-05-27-stage-5-install-sh-k1-verification.md:24) document the pre-withdrawal era; the design spec + plan (docs/superpowers/specs/2026-05-31-one-click-installer-design.md:101, docs/superpowers/plans/2026-05-31-one-click-installer.md:466) document the withdrawal itself — consistent.
  • Observation, not fixed: packages/core/skills/classify-work.test.ts:19 attributes TaskMaster complexity tiers to «SSOT fix(hooks): add permission for end-of-turn-reminder Stop hook #72», but fix(hooks): add permission for end-of-turn-reminder Stop hook #72 is OpenHands (prior-art-evaluations.md:142); TaskMaster is fix(hooks): aggregate full assistant turn for end-of-turn reminder #73 (prior-art-evaluations.md:143) — stale ID in a test comment.

@artyhoo
artyhoo enabled auto-merge (squash) June 10, 2026 18:17
@artyhoo
artyhoo merged commit 3220e18 into staging Jun 10, 2026
25 checks passed
@artyhoo
artyhoo deleted the worktree-oci-s4-ssot branch June 10, 2026 18:24
artyhoo added a commit that referenced this pull request Jun 10, 2026
…non-fast-forward diagnosis (#455)

Modes: every blob's tree entry hardcoded mode 100644, silently stripping
the exec bit off executables and mispushing symlinks. Now the mode is
resolved per path — git index (ls-files --stage) → HEAD (ls-tree) →
filesystem bits fallback — so 100755 and 120000 survive; a symlink blob
carries the link target string; gitlink/tree modes (160000/040000) are
rejected loudly instead of being pushed as corrupt 100644 blobs.

Ref edge: the branch-update PATCH {force:false} now diagnoses a 422
non-fast-forward explicitly — shows the moved remote tip and the new
commit's actual parent, suggests re-run or a fresh --branch, and states
it will NOT auto-force — instead of dying on a raw gh error. Found live
while dogfooding this push: branch existence must key off gh's exit
code, not stdout — on 404 `gh api --jq` prints the error body to stdout,
so a non-empty-stdout check takes the PATCH path for absent branches
(pinned in the test stub).

Incidents (one-click-installer umbrella, 2026-06-10): the S4 pushes
(#449-#451) required hand-editing the tree payload to preserve the
repo-root `setup` script's 100755 mode, and the S5 push (#452) needed
manual hand-holding of the force:false ref PATCH. Both adaptations are
now parameterized and pinned by a paired-negative test
(tests/dispatcher/harvest-via-api.test.sh — stubbed gh, no network),
wired into audit-self.yml#principles-meta-tests.

Prior-art: skipped — hardening + test for an existing helper, no new capability.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant