| name | first-tree-github-scan |
| version | 0.4.0-alpha.1 |
| cliCompat | {"first-tree":">=0.4.0 <0.5.0"} |
| description | Agent skill for handling a single GitHub notification spawned by the First Tree github-scan daemon. Use when an agent needs to decide how to tag, comment on, escalate, or hand off a PR, issue, review request, mention, or CI event, and when it needs to choose between handling on GitHub or routing to `first-tree-sync` / `first-tree-write`. This is the agent behavior spec, not the human CLI operations guide for the daemon — for that, run `first-tree github scan --help`. |
First Tree Github Scan
Read these first:
../first-tree/SKILL.md../first-tree/references/functions.md../first-tree/references/anti-patterns.md
What This Skill Does
This skill is loaded by the daemon every time a new GitHub notification
arrives and a per-task agent is spawned. The agent's job is to:
- classify the notification
- set the right
github-scan:* label and post any necessary comment
- choose exactly one of four routes (
reply / human / sync / write)
- emit a
GITHUB_SCAN_RESULT: line so the daemon can dispatch
downstream work
This is not the operator-facing daemon guide. Humans running
first-tree github scan install / start / doctor should use the CLI
help, not this skill.
When To Use This Skill
| Use this skill | Use a different skill |
|---|
| The daemon spawned an agent for one notification | A user wants to onboard a repo → first-tree-onboarding |
| You need the tag table or comment template | A user wants to audit drift → first-tree-sync |
| You need the route taxonomy | A user wants to write tree content from a source → first-tree-write |
| You need to decide whether to involve a human | An operator is debugging the daemon — use first-tree github scan --help |
The Four Tags
github-scan:new — initial state; daemon also restores via auto-revert
github-scan:wip — agent has started real work
github-scan:human — a human must step in
github-scan:done — finished, no further action
Exactly one github-scan:* label per item at any time. Full lifecycle
and gh calls in references/tags.md.
The Four Routes
reply — handle on GitHub: labels / review / comments
human — set github-scan:human and stop
sync — load first-tree-sync SKILL.md and continue under that skill
write — load first-tree-write SKILL.md and continue under that skill
Pick the most specific route that fits — see
references/route-taxonomy.md for the
selection algorithm and worked examples. Required final-line format:
GITHUB_SCAN_RESULT: status=<handled|skipped|failed> route=<reply|human|sync|write> summary=<one-line>
Hard Rules
- Do not edit the tree from this skill. Routing to
sync or write
is how tree changes happen. Direct tree writes here are forbidden.
- Do not invent new
github-scan:* tags. The four are the contract.
Missing-label situations route to human with a skipped status.
- One tag at a time. When flipping state, remove the previous label
in the same operation.
- Set the terminal tag before posting the final comment. Race
conditions with the auto-revert otherwise confuse the daemon.
- Never reply with code diffs or chain-of-thought. Comments are
short and decision-grounded — see
references/comments.md.
- Include the disclosure sentence verbatim on every public comment.
If quoting a user-requested string would change the disclosure, post
the disclosure separately.
- Route
human only on the four triggers in
references/when-human.md. Reassurance and
size are not triggers.
References
- tags.md — the four-tag protocol, lifecycle
diagram,
gh invocations, auto-revert mechanic, missing-label
fallback
- route-taxonomy.md — route definitions,
selection algorithm, worked examples,
GITHUB_SCAN_RESULT format
- comments.md — required fields, tone, length,
per-route templates, disclosure handling, what comments must never
contain
- when-human.md — the four triggers (new
decision / owner unclear / cross-domain impact / rules cannot pick)
and anti-triggers
