docs

name	docs
description	Propose-review-apply flow for BMad artifacts (PRDs, UX specs, architecture, briefs). All applies require --i-approve and produce commits with Proposed-By/Approved-By trailers.

docs

Safety gate for BMad-document edits. Steward proposes; human (Luci) applies.

Steward produces a unified diff + rationale and calls docs propose. The diff lands as a draft patch in .pi/.doc-drafts/. Luci reviews it with docs list and docs show <id>, then runs docs apply <id> --i-approve to land it as a structured commit. Rejected drafts are preserved under .pi/.doc-drafts/.discarded/ for audit.

Subcommands

propose

# Read unified diff from stdin; require rationale explaining why
bun run .pi/skills/docs/bin/docs.ts propose docs/PRD.md --rationale="add onboarding section"

• Scope check: <path> must start with docs/, specs/, or specs/briefs/. Anything else is refused with a clear error.
• Validates that stdin is a well-formed unified diff (must contain --- and +++ headers).
• Writes .pi/.doc-drafts/<ISO-timestamp>-<slug>.patch (mode 0600).
• Prints the draft ID to stdout.

list

bun run .pi/skills/docs/bin/docs.ts list

Lists all pending drafts as a table: id | target | proposed_at | rationale (first 60 chars). Prints "no pending drafts" when the queue is empty. Does not show discarded or applied drafts.

show

bun run .pi/skills/docs/bin/docs.ts show 2026-04-24T14:02:11Z-prd-md

Prints the full patch file (rationale block + unified diff) to stdout. Exits non-zero if the ID is not found.

apply

# Dry-run (default): preview what would be applied
bun run .pi/skills/docs/bin/docs.ts apply 2026-04-24T14:02:11Z-prd-md

# Approved apply: patch, stage, commit
bun run .pi/skills/docs/bin/docs.ts apply 2026-04-24T14:02:11Z-prd-md --i-approve

Without --i-approve: prints the full diff + "NOT YET APPLIED (rerun with --i-approve)". Exit 0.

With --i-approve:

1. Verifies working tree is clean (git status --porcelain must be empty).
2. Applies the patch with git apply --index.
3. Reads .pi/.honcho-state.json to determine if a SpecSafe slice is open.
4. Commits with structured trailers:

   docs: <rationale>
   
   <rationale>
   
   Proposed-By: steward
   Approved-By: luci
   Spec-Slice: <sliceId>     (omitted if no slice open)
   Rationale-From: <draft-id>
   

5. Moves the patch file to .pi/.doc-drafts/.applied/<id>.patch (forensic retention).

discard

bun run .pi/skills/docs/bin/docs.ts discard 2026-04-24T14:02:11Z-prd-md

Moves the draft to .pi/.doc-drafts/.discarded/<id>.patch. Creates .discarded/ if absent. Exits non-zero if the ID is not found.

Allowed paths

Only paths beginning with one of these prefixes are accepted by propose:

Prefix	Contains
docs/	PRDs, UX specs, architecture docs
specs/	SpecSafe spec slices
specs/briefs/	Linear-ticket briefs produced by Steward

Any other path — including README.md, src/, etc. — is refused with:

docs skill is scoped to BMad artifacts (paths must start with docs/, specs/, or specs/briefs/)

Typical workflow

# 1. Steward proposes a diff (run by Steward agent):
git diff HEAD docs/PRD.md | bun run .pi/skills/docs/bin/docs.ts propose docs/PRD.md \
  --rationale="add section 4 — onboarding flow per CUR-42"

# 2. Luci reviews:
bun run .pi/skills/docs/bin/docs.ts list
bun run .pi/skills/docs/bin/docs.ts show 2026-04-24T14:02:11Z-prd-md

# 3a. Luci approves:
bun run .pi/skills/docs/bin/docs.ts apply 2026-04-24T14:02:11Z-prd-md --i-approve

# 3b. Luci rejects:
bun run .pi/skills/docs/bin/docs.ts discard 2026-04-24T14:02:11Z-prd-md

Storage

All draft files live under .pi/.doc-drafts/ (gitignored, not committed to the repo):

.pi/.doc-drafts/
  2026-04-24T14:02:11Z-prd-md.patch     # pending
  .applied/
    2026-04-24T13:50:00Z-ux-spec-md.patch  # landed
  .discarded/
    2026-04-24T13:30:00Z-arch-md.patch     # rejected

No Co-Authored-By

Commits created by docs apply carry no Co-Authored-By trailer. The commit represents a human decision (Luci's approval) on a Steward-proposed change. The Proposed-By: steward and Approved-By: luci trailers make the authorship chain explicit without attributing to any AI agent.