| name | docs-all |
| description | Run all documentation skills (docs-readme, docs-diagrams, docs-improve) in one
command, dispatching each as a sub-agent in an order chosen per run (with a
documented default precedence as fallback), and return a single consolidated
report. Use to refresh the whole doc set at once.
|
All-in-One Documentation Refresh
Orchestrate the three existing documentation skills in a single pass instead of
running them by hand. Each runs as an independent sub-agent; results are merged
into one report.
When to use
- You want to refresh README, architecture diagrams, and the Diataxis docs audit
together after a meaningful change.
- You don't want to remember the right order to run them in.
Task
-
Decide the order for this run. Inspect what changed (recent diff / the
target path) and choose an order, honoring the one hard dependency:
docs-improve runs last (its quality audit should see the README and
diagram updates the other two produce).
- Default precedence (fallback when no strong signal):
docs-readme → docs-diagrams → docs-improve.
- Signal-based adjustments:
- Changes concentrated in architecture/module structure or imports →
prioritize
docs-diagrams early.
- Changes mostly in prose/onboarding → prioritize
docs-readme early.
- Either way, keep
docs-improve last.
- If the user passed
--order a,b,c, use it verbatim (still keep the
dependency in mind and warn if it is violated).
-
Dispatch each docs skill as a sub-agent using the Task tool, one per
skill, passing through the target path. Run them per the order from step 1.
Independent skills may run concurrently; docs-improve waits for the others.
-
Continue on failure. If one sub-agent fails, capture its error and still
run the remaining skills — never abort the whole run because one failed.
-
Emit one consolidated report:
docs-all report
Order: <a> → <b> → <c> (reason: <signal | default fallback>)
- docs-readme : success — <1-line summary>
- docs-diagrams : success — <1-line summary>
- docs-improve : failed — <error>
The report MUST state the order used, why that order was chosen, and each
sub-skill's outcome (FR-010/FR-011).
Notes
- This skill writes no files itself and runs no shell script — it is pure
sub-agent orchestration over
docs-readme, docs-diagrams, docs-improve.
- If one of the three skills is unavailable, run the others and report the gap.
- Verification scenario: see
specs/002-new-agent-skills/quickstart.md (docs-all
section) for how to confirm ordering, rationale, and failure-surfacing.