Skip to content

feat(ac-store): salvage ACs from closed PRs #113 and #1 (BO-2000, TQ-200)#238

Open
urlmonitor wants to merge 1 commit into
mainfrom
salvage/pr-cleanup-followups
Open

feat(ac-store): salvage ACs from closed PRs #113 and #1 (BO-2000, TQ-200)#238
urlmonitor wants to merge 1 commit into
mainfrom
salvage/pr-cleanup-followups

Conversation

@urlmonitor

Copy link
Copy Markdown
Owner

Summary

Re-authors the genuinely-valuable content from two PRs that were closed as unmergeable (stale, conflicting, and colliding on reassigned AC IDs), under fresh collision-free IDs via the canonical PO v3 → BA v3 → IT PO v3 flow (ADR-012).

All ACs are readiness: draftqueued for review; they will not be picked up by /build-ac until promoted to approved.

BO-2000-self-protecting-build (salvages #113)

Original #113 collided on the reassigned BO-1500 id. Re-authored as a new BO-2000 tree, explicitly disambiguated from BO-1700 (hook execution in worktrees) and BO-1500 (safe AC-authoring delivery):

  • BO-2000a (KI-2) — self-hosting non-regression claims must be backed by a recorded command + captured zero-new-violations output; bare "verified" claims are rejected by the AC-fulfillment gates (ac-fulfillment-gate 11.7 / ac-validator 11).
  • BO-2000b (KI-3) — builds must never run over an active worktree root (it clobbers tracked build-output and corrupts the diff); deploy is exercised against a disposable --target-dir outside the worktree; dual-tree files are edited in templates/ and compared, never build-synced.

TQ-200-tooling-coverage-recovery (salvages #1)

Recovers unit-test coverage for 9 currently-untested package-internal tooling modules: registry_validator (core + spawn bidirectionality), registry injection (injection_builders + template_compiler), package_audit, selection_criteria_evaluator, generate_agent_diagram, prioritize, scan_secrets.

⚠️ Open questions for review (from IT PO enrichment)

Several TQ-200 L2 criteria describe behavior that does not exist in the current code — decide whether to amend to pure coverage-recovery or expand scope:

  • prioritize.py has no parked-ticket handling (_is_done only knows done/deferred).
  • registry_validator has no missing-field / duplicate-id validation (missing id raises KeyError).
  • selection_criteria_evaluator DSL has no negation operator.

And for BO-2000a: the evidence-bundle location is unspecified (sign-off comment vs. artifact file vs. AC field) — needs a BA decision before the gate rule can be written.

Test plan

  • AC schema validation passes (check_ac_schema.py, exit 0).
  • No AC-ID collisions (full-store scan loads all trees without duplicate-id error).
  • Both trees under the 20-AC tree cap; parent covered_by lists wired.
  • No code changes — AC-store only.

🤖 Generated with Claude Code

Re-authors the genuinely-valuable content from two closed-as-unmergeable PRs
under fresh, collision-free AC IDs via the PO v3 -> BA v3 -> IT PO v3 flow.

BO-2000-self-protecting-build (from #113, which collided on the reassigned
BO-1500 id):
  - BO-2000a (KI-2): self-hosting non-regression claims must be backed by a
    recorded command + captured zero-new-violations output; bare 'verified'
    claims are rejected by the AC-fulfillment gates.
  - BO-2000b (KI-3): builds must never run over an active worktree root;
    deploy is exercised against a disposable --target-dir outside the worktree.

TQ-200-tooling-coverage-recovery (from #1): recover unit-test coverage for 9
currently-untested package-internal tooling modules (registry_validator,
package_audit, selection_criteria_evaluator, injection/template_compiler,
generate_agent_diagram, prioritize, scan_secrets).

All ACs are readiness: draft (queued for review; will not auto-build).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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