← ClaudeAtlas

add-cross-sf-consistencylisted

Use when judging contract consistency across an epic's subfeature plans — the five-dimension rubric, dedupe rules, and finding routes.
brabos-ai/code-addiction · ★ 9 · AI & Automation · score 76
Install: claude install-skill brabos-ai/code-addiction
# add-cross-sf-consistency — Cross-Subfeature Consistency Rubric <!-- uses: - skill: add-doc-schemas - skill: add-plan-review - skill: add-qa - mention: add-code-review - skill: add-doc-schemas/references/review.md - mention: @qa-agent - agent: backend-agent - agent: consistency-agent - agent: database-agent - agent: frontend-agent - agent: plan-reviewer-agent - agent: reviewer-agent - agent: ux-agent - command: /add.build - command: /add.plan - mention: /add.review - mention: add-review-discipline - mention: @fix-agent --> ## Overview The rubric `consistency-agent` loads before it reports. It is dispatched twice per epic run: a **FULL pass** by `/add.plan` after each subfeature's `plan.md` is consolidated and reviewed (comparing it against every already-converged sibling), and one **DELTA pass** by `/add.build`, from its `## Final Review` on the epic's last subfeature, before that subfeature's checkpoint (re-checking only what changed since the last verdict). The agent compares declared contracts **document against document** — `plan.md`, `about.md`, `design.md` — and never opens source code; code-level review stays `@reviewer-agent`'s job. This produces no new persisted artefact of its own. A `FULL`-pass finding is applied straight into the subfeature's own `plan.md`; a `DELTA`-pass finding joins `/add.build`'s final review, whose verdict is the `Final review:` ledger line. There is no `consistency-validation-NNN.md` and none should ever be invented — see Where Findings