The prior commit pointed bmad-code-review at bmad-quick-dev's
deferred-work-format.md via a relative path (`../bmad-quick-dev/...`), which
reaches across a skill boundary and violates the PATH-05 encapsulation rule
(tools/skill-validator.md) — fragile when the skills are installed
independently. Ship a local copy inside bmad-code-review and reference it
locally instead:
- add bmad-code-review/deferred-work-format.md (identical to bmad-quick-dev's,
plus a one-line "keep the two in sync" note).
- automation-mode.md rule 7: reference `./deferred-work-format.md`.
- steps/step-04-present.md: reference `../deferred-work-format.md`.
bmad-quick-dev is unchanged. validate:skills / validate:refs / lint:md /
format:check all pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Resolves the augmentcode review findings on PR #2494, tightening the
BMAD_AUTO_MODE contract so unattended runs behave consistently:
- code-review automation-mode.md: harmonize the result.json `clean`
definition with rule 8 (zero actionable findings AND spec status
set to `done`), and correct the `dismissed` count description
(recorded to the Review Ledger, not dropped).
- code-review step-04 §1: the clean shortcut now keys off actionable
findings only and, in auto mode, still writes the Review Ledger when
findings were dismissed before jumping to §6 — otherwise set-aside
dismissals never reach the ledger and the next cycle re-litigates them.
- code-review: give explicit relative paths to bmad-quick-dev's
deferred-work-format.md (rule 7 and step-04) so unattended agents can
locate it.
- code-review step-01: ground the auto-mode note in rule 1 (no HALT
anywhere in the step) instead of enumerating a partial skip list.
- quick-dev deferred-work-format.md: drop the dead
`bmad-auto-sweep/migration-mode.md` reference (that file ships with the
orchestrator, not BMAD-METHOD); describe the sweep migration as an
external orchestrator-side process.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add an opt-in automation mode (gated on BMAD_AUTO_MODE=1) so both skills can
run unattended under the bmad-auto orchestrator: checkpoints resolve via
decision tables instead of halting for human input, deferred work and review
findings are written in machine-readable formats, and the code-review skill
gains a static prefilter, prior-cycle ledger, and code-verification triage pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
pbean