el-ai-brieflisted
Install: claude install-skill threatroute66/EL
# el-ai-brief — Claude Code fulfilment of an EL executive brief
EL's `executive_ai.py` chooses among three transports:
1. **Cache hit** (no work for this skill).
2. **Direct API** (`ANTHROPIC_API_KEY` set — EL calls the Anthropic
SDK in-process; this skill not needed).
3. **Claude Code skill** (no API key, but EL detected `CLAUDECODE=1`
or the operator passed `--defer-ai-brief`). EL writes a request
file at `reports/_ai_brief_request.json`; this skill fulfils it.
**You are path 3.** The request file's `trigger` field tells you
which sub-path fired (`claude_code_session` vs
`explicit_defer_flag`). Both produce the same response shape.
## When to fire
Fire when **any** of the following is true:
- The user explicitly invokes `/el-ai-brief` (with or without a case-id arg).
- A new `_ai_brief_request.json` file appears under any
`/opt/EL/cases/<id>/reports/` directory (e.g. after an `el investigate`).
- A new `_combined_ai_brief_request.json` file appears under any
`/opt/EL/cases/_combined/<bundle>/` directory (after an
`el combined-report`). Same shape as per-case requests but carries
`brief_kind: "combined_executive"` and a different output path.
If the user hasn't asked for it and there are no pending requests on
disk, do nothing — this skill is a worker, not a watchdog.
## Procedure
1. **Discover pending requests**.
```bash
find /opt/EL/cases -maxdepth 4 \
\( -name "_ai_brief_request.json" \
-o -name "_combined_ai_brief_r