From rad-repo-manager
This skill should be used when the user says "repo-align", "align the repo", "clean up the repo", "check for drift", "is the repo in good shape", "find contradictions", "what's gotten messy", "doc cleanup", "are my docs consistent", "tidy up the docs", or when startup flagged drift. The deep, opt-in alignment pass: finds documentation drift (duplicate authorities, stale docs, contradictions, scope-creep, loose/misplaced docs, broken read paths) and proposes fixes interactively. Proposes before acting on every judgment call; preserves history with git mv; never silently merges, deletes, or archives.
How this skill is triggered — by the user, by Claude, or both
Slash command
/rad-repo-manager:repo-align [--report-only][--report-only]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
The deep, opt-in pass. Go deep enough to find documentation drift, surface it in plain
The deep, opt-in pass. Go deep enough to find documentation drift, surface it in plain English, and propose one disposition per judgment call — always confirmed before acting. Never silently merge, delete, archive, or rewrite.
This can take a minute. The fast everyday skills are startup/wrapup; this is the
periodic alignment. With --report-only, do steps 1–3 (find and present) and stop —
propose no changes.
Active read path (the four core docs, declared in AGENTS.md):
AGENTS.md — operating manual / how we workdocs/prd.md — durable product authoritydocs/plan.md — the plan (owned by /rad-planner:plan)docs/handoff.md — the short resume snapshotConditional (read only when the task touches them): docs/design.md,
docs/reference/* (closed catalog), docs/README.md. docs/archive/* is history, not
read by default. READMEs explain the repo to humans — they are not agent
instruction authority.
These are the drift signals — pure-stdlib, advisory. Capture each JSON; they surface
candidates, not verdicts. Your judgment decides which are real. (Use python3,
or python on Windows.)
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/repo-scan.py . --json --no-record # floating/misplaced docs, active-set growth, AGENTS.md bloat
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/doc-freshness.py . --json # stale handoff; prd/plan unchanged while code churned
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/doc-contradiction.py . --json # plan builds something the PRD lists as a non-goal (scope-creep)
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/doc-redundancy.py . --json # the same fact stated in two active docs
python3 ${CLAUDE_PLUGIN_ROOT}/scripts/audit-user-content.py . --json # AGENTS.md orphan terms + dead paths
(If Python 3 is unavailable, do the same checks by reading the docs directly and say the mechanical pass was skipped.)
Read AGENTS.md, docs/prd.md, docs/plan.md, docs/handoff.md. Then search for
repo-management and doc files that may be drifting:
CLAUDE.md, GEMINI.md, README.mddocs/status.md, docs/roadmap.md, docs/implementation-plan.md (off-model — should not exist)docs/design.md, docs/reference/*, docs/archive/**.md outside the expected locations; docs/inbox/* (a retired tier — file its contents out)AGENTS.md/CLAUDE.md/GEMINI.md (likely confuse agents)Identify: duplicate authorities, stale docs, pointer chains, missing core docs, contradictory read paths, docs that belong in archive or the reference catalog, root-README content that has become stale status, and off-model status/roadmap docs.
The question that matters for fuzzy findings: does this mean a decision is needed? A prd↔plan contradiction usually does — frame it as a decision for the user, not an error to silently patch.
# Repo alignment
## Needs a decision
- The plan builds offline mode, but the PRD lists offline as a non-goal. Which wins?
## Active-doc conflicts / redundancy
- `docs/prd.md` and `AGENTS.md` both define the validation steps — pick one home.
## Off-model docs
- `docs/status.md` exists — this model has no status doc; fold its live bits into docs/handoff.md, archive the rest.
## Loose / misplaced docs
- `smoke-2026-05-30.md` (repo root) — a smoke-test report sitting loose.
- `docs/inbox/brainstorm-routing.md` — inbox is retired; file this out.
## Pointer / read-path problems
- `AGENTS.md` links `docs/old-architecture.md` — file doesn't exist.
## Missing core docs
- `docs/handoff.md` is absent.
## Suggested actions
[per-item dispositions below]
No jargon, no validator names in the user-facing report — describe the problem and the fix in words a non-coder follows.
For each drifting or misplaced doc, propose ONE disposition and ask before acting (AskUserQuestion, or a clear yes/no per item):
docs/plan.md or docs/reference/lessons-learned.md, then archive the original.docs/reference/ catalog slot (decision-log, architecture, api-contracts, commands, lessons-learned, testing) or to top-level docs/design.md.git mv into docs/archive/, add the archive banner.Rules:
git mv, never delete-and-recreate.CLAUDE.md/GEMINI.md as thin @AGENTS.md pointers — if one has grown its own content, propose folding that content into AGENTS.md and restoring the shim.When a finding implies a change to a durable doc the manager doesn't own — docs/prd.md
(product behavior) or docs/reference/decision-log.md (a decision) — the user owns the
decision, not the typing. Draft the exact edit (precise wording, old → new) and ask
per doc via AskUserQuestion: apply / skip / let me reword. Apply only on an explicit
"apply" for that specific edit; a skip means hands off, restated in one line at the end.
Never bundle user-owned edits into a blanket OK. The manager drafts and flags; the user
owns product and decisions.
docs/prd.md or the decision log without an explicit per-edit "apply" — it drafts the exact edit and asks.docs/status.md, docs/roadmap.md, docs/implementation-plan.md,
loose root status/handoff/audit docs, or folder-specific agent files.${CLAUDE_PLUGIN_ROOT}/references/doc-model.md — tiers, active core, reference catalog, filing${CLAUDE_PLUGIN_ROOT}/scripts/ — the four drift-signal scansnpx claudepluginhub radorigin-llc/rad-claude-skills --plugin rad-repo-managerProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.