| name | project-health |
| description | Run the read-only diagnostics — doctor, health, status, followups, product-consistency, audit — and read their combined output as a pre-flight check. Use when asked "is the project healthy", "run diagnostics", "what's blocked", or "check corpus health" before starting other work. Writes nothing; capturing findings as a request is /review-project's job, not this one. |
Project health diagnostics
Run the corpus's own read-only diagnostics and read them together as one
pre-flight check. This skill produces no document — it's the thing you run
before deciding whether /review-project or /implement-task is next.
Hard rules
- This skill writes nothing. No request, no backlog item, no doc edit.
If a finding here turns out to deserve a captured request, that's
/review-project's job — hand off to it, don't duplicate its capture step
here.
- Report what the tools actually said, not a summary that drops the
specific doc paths and remediation commands they printed. An agent reading
your output next needs the exact command to run, not just "there are
issues."
- Don't fix anything you find.
doctor's remediation lines and
audit's repair-command suggestions are for /closeout-repair or the
operator to act on, not for you to apply silently mid-diagnostic.
Recipe
Run each, in this order (cheapest structural check first, broadest last):
logics-manager doctor
logics-manager health
logics-manager status
logics-manager followups
logics-manager product-consistency
logics-manager audit --group-by-doc
Add --format json to any of them if you need to correlate findings
programmatically rather than read prose.
Reading the combined output
doctor failing on missing_schema_version across old docs is usually
pre-existing debt, not a new break. Check whether the flagged docs
predate the schema-version convention before treating it as urgent —
doctor's own remediation line tells you the exact fix either way.
health's issue signals: 0 doesn't mean the corpus is clean — it
means the specific signals health tracks are clean. Cross-check against
audit, which runs the full traceability/consistency pass health
doesn't.
status's "Continue or finish N active task(s)" is the fastest way to
find in-flight work before starting something new — check it before
assuming a fresh task is needed.
followups gives you a ready flow new request command per finding —
if you're about to write one by hand from an audit finding, check
followups first; it may have already built the exact command.
- A blocking
audit finding always outranks a doctor/health warning.
If time is short, read audit's blocking issues first, then decide
whether the rest is worth a full pass.