| name | document-all-nodes |
| description | Find every node in @rosepetal/node-red-contrib-image-tools that lacks
documentation (no docs/<cat>/<name>.md or no assets/nodes/<cat>/<name>-demo.gif)
and run /document-node on each, sequentially. Reports a final table of
successes and failures. Pass --include-existing to re-document every
registered node, regardless of current state.
|
| when_to_use | Trigger phrases: "document all nodes", "document missing nodes", "fill in
the missing docs", "regenerate all node docs". No required arguments.
|
| argument-hint | [--include-existing] |
| allowed-tools | Read Bash Skill |
| disable-model-invocation | false |
/document-all-nodes
Batch wrapper that runs /document-node for every orphan node in this repo.
Procedure
-
Discover orphans. Read package.json and list rp-* keys under node-red.nodes. For each, derive the bare name and category (from the JS path). A node is an "orphan" when either docs/nodes/<cat>/<name>.md is missing or assets/nodes/<cat>/<name>-demo.gif is missing. Use Bash to run find / test -f rather than reading directories manually.
With --include-existing, treat every registered node as orphan.
-
Print a plan. Show the user the list you'll process, in order. Sort by category, then alphabetically.
-
Process sequentially. For each orphan, invoke /document-node <name> via the Skill tool and wait for the result before moving on. Why sequentially: the GIF recorder needs exclusive access to the headless browser page; running concurrent doc runs would race on Node-RED state.
-
Per-node bookkeeping. Track success/failure for each. On failure, capture the failure reason but continue with the next node — don't abort the whole batch.
-
Final summary. Print a Markdown table:
| Node | Category | .md | GIF | Status |
|---|
| folder-in | io | ✅ written | ✅ 312 KB | OK |
| add-bbs | blend | ✅ written | ❌ failed (size) | Reviewer flagged |
| … | … | … | … | … |
Then a one-line totals row: "X/Y nodes documented end-to-end. Z need manual review."
Hard rules
- No parallelism. Sequential only.
- Don't skip Phase 0 for each
/document-node run — the validator is idempotent and cheap; let it confirm the environment didn't drift mid-batch.
- Don't keep going if Phase 0 fails for any node — that's an environment problem affecting the whole batch. Stop and surface the validator's note immediately.
- English only in the final report and any per-node messages.
