create-skill

> **${var}** — A natural-language description of the skill to create. **Required.** Example: `"monitor Hacker News for AI papers and send a summary"` or `"track gas prices on Ethereum and alert when below 10 gwei"`.

<!-- autoresearch: variation B — sharper output via PR-first workflow + quality enforcement + exit taxonomy + new-secret guard -->

If `${var}` is empty, exit `CREATE_SKILL_NO_VAR`:
```bash
./notify "create-skill aborted: var empty — pass a description e.g. \"monitor X for Y\""
```
Then stop.

Today is ${today}. Your task is to generate a complete, production-ready skill from `${var}`, score it against a quality bar, and ship it as a PR — **never commit directly to `main`**.

## Steps

1. **Parse the request.** Extract from `${var}`:
   - Core action verb (monitor, fetch, generate, analyze, alert, track, scan, etc.)
   - Data source(s) — APIs, websites, RSS, on-chain, GitHub, etc.
   - Output format — notification, article, file, PR, dashboard, etc.
   - Configurable parameter(s) the new skill will accept via its own `${var}`
   - Suggested cadence (daily, hourly, weekly, on-demand)

   Save a one-paragraph structured request summary; you'll use it in the PR body.

2. **Duplicate detection (deep — not just `ls`).** Find functional overlap, not just name collision.
   ```bash
   keywords=$(echo "${var}" | tr '[:upper:]' '[:lower:]' | grep -oE '[a-z]{4,}' \
     | grep -vE '^(send|with|from|that|this|when|each|into|over|some|like|just|than|then|also|will|have|been|using|monitor|track|fetch|alert)$' \
     | sort -u)
   for kw in $keywords; do
     grep -liE "$kw" skills/*/SKILL.md | head -5
   done
   ```
   Read the top 3 candidates fully. For each, judge: does it already do this? Could the request be solved by running an existing skill with a different `var=`?
   - **Near-duplicate exists** → exit `CREATE_SKILL_DUPLICATE`. Notify with the existing skill name and a one-line suggestion ("use existing `{skill}` with `var={...}` instead"). Stop.
   - **Functionally adjacent** → design the new skill to complement (different angle/cadence/output). Document the boundary in the PR body.

3. **Research the data sources.** For every API or data source the new skill needs:
   - **WebSearch** for the current API documentation. Cross-check against a secondary source when feasible (a recent GitHub repo using it, an official changelog, or a Stack Overflow answer dated ≥2026) to confirm the endpoint isn't deprecated.
   - **WebFetch** the canonical docs URL — record it as a comment in the SKILL.md and in the PR body's "Sources researched" section.
   - Identify exact endpoints, required headers, auth scheme, response schema, and rate limits.
   - Note every required environment variable / API key.
   - Determine fallback strategy when an optional API key isn't set (WebSearch / WebFetch / cached data / public endpoint).

   **Research bar (soft):** at least one confirmed source URL or exemplar (a working docs page or a public repo using the API). If none, do **not** hard-abort — log `CREATE_SKILL_INSUFFICIENT_RESEARCH`, ask the operator via `./notify` with what was tried and why each source failed, and stop. The operator can re-dispatch with a clearer prompt or a source hint.

4. **New-secret guard.** Secrets values are never inspectable from the workflow — only names are listed. Use `gh api repos/:owner/:repo/actions/secrets --jq '.secrets[].name'` to read the **names** of secrets already configured (this endpoint returns names only, never values). Cross-reference with env-var usage in `aeon.yml` and existing workflows. For each env var the new skill needs:
   - **Name present** → continue.
   - **Name missing** → record as `NEW_SECRET_REQUIRED`. The generated skill **must** gracefully degrade or skip when the secret is absent (no hard crash). Add a `### Required secrets` section to the PR body listing what the operator must add to GitHub Actions secrets before enabling.

   If the secret has no graceful fallback, the generated skill's step 1 must do:
   ```bash
   if [ -z "$VAR" ]; then ./notify "{skill} skipped: VAR not set"; exit 0; fi
   ```

5. **Design the skill.** Decide:
   - **Skill name** — lowercase, hyphenated, 2-3 words max (e.g., `gas-alert`, `hn-papers`). Must not collide with any existing entry under `skills/`.
   - **Description** — one sentence, starts with a verb, ≤90 chars.
   - **Tags** — pick from: `content`, `crypto`, `dev`, `meta`, `news`, `research`, `social`. Max 3.
   - **Variable behavior** — what `${var}` controls; what happens when empty (sane default OR clean abort with notify).
   - **Steps** — 4-8 numbered, following the standard pattern: read context → fetch/search → process/analyze → write output → log → notify.
   - **Schedule suggestion** — choose a cron slot. Read existing schedules in `aeon.yml`; avoid co-scheduling at the same minute as heavy skills (article, repo-scanner, deep-research, telegram-digest) unless the new skill is lightweight (<30s expected). Prefer a `:30` minute offset if the natural hour is already crowded.
   - **Model** — default `claude-sonnet-5`. Pick `claude-haiku-4-5-20251001` if the skill is high-frequency aggregation/digestion (cost optimization), or `claude-opus-4-8` if it needs the strongest reasoning. Document the choice in the PR body.
   - **Category** — the pack the skill joins. Pick one: `research` `dev` `crypto` `onchain-security` `social` `productivity` `meta`. (`core` and `fleet` are curated in `packs.config.json`, not chosen here.) If none fits, omit it and the skill lands in the **Lab** catch-all for later triage. See `docs/skill-packs.md`.

6. **Write the SKILL.md draft** at `skills/{skill-name}/SKILL.md` with this exact structure:

   ```markdown
   ---
   name: {skill-name}
   description: {One-sentence description starting with a verb}
   metadata:
     title: {Display Name}
     category: {category}
     var: ""
     tags:
       - {tag}
   ---
   > **${var}** — {What the variable controls}. {If-empty behavior