email_triage

Files (2)

• SKILL.md
• skill.py

SKILL.md (first lines)

# email_triage

Reads the last N email leaves from the Memory Tree (`domains/email/<account>`)
that Phase F's Gmail adapter has already filed. Asks the brain to
classify each one into:

- **URGENT** — needs operator attention now
- **ACTION_REQUIRED** — needs operator action soon
- **INFORMATIONAL** — useful context, no action
- **TRANSACTIONAL** — receipts, shipping, automated confirmations
- **SECURITY_ALERT** — auth / login / token rotation
- **NOISE** — newsletters, marketing, mass mailers

Saves a single Markdown report at
`domains/email/_triage/triage_<yyyy-mm-dd>.md` with per-email rows,
per-category counts, and a "next step" block of the URGENT and
ACTION_REQUIRED items.

## CLI

```
python3 ~/aria/skills/email_triage/skill.py
       [--since-hours N]      # default 24
       [--max-emails N]       # default 30
       [--account ADDR]       # filter to one account
```

## Behaviour

* Reads from the local Memory Tree, NOT the Gmail API. Phase F's
  `gmail` auto-fetch adapter already files emails as leaves.
* Empty tree / brain unreachable → graceful fallback, exit 0,
  verifier marks `[UNVERIFIED]`, brain narrates the truth.
* Read-only — no Gmail label moves, no replies, no archival.
  Drafts are a future Phase L.2.

## Verifier

* `exit_code == 0`
* `output_contains "triage saved at"` (only matches on the success
  path; the fallback messages don't contain this token, so the
  brain can quote them as `[UNVERIFIED]` truthfully)

## Trust tier

`1` — pure read + summarize. No outbound, no destructive write.

From GamerX3560/Aria-V12, MIT.

More from GamerX3560/Aria-V12