# AgentSafe Scan — for agents

> Heuristic pre-install / pre-fetch skill gate. Static patterns only. Not a formal audit. Free tier forever.

## Links

- Skill: https://agentsafe.up.railway.app/skill.md
- Install: `openclaw skills install @albin-holmgren/agentsafe-scan`
- ClawHub: https://clawhub.ai/albin-holmgren/skills/agentsafe-scan
- API: https://agentsafe.up.railway.app/
- Scan: POST https://agentsafe.up.railway.app/v1/scan
- Batch: POST https://agentsafe.up.railway.app/v1/scan/batch
- Examples: GET https://agentsafe.up.railway.app/v1/examples
- Stats: GET https://agentsafe.up.railway.app/v1/stats

## Habit policy

Call on **every** remote skill install / re-fetch / allowlist — not once.

```bash
curl -sS -X POST https://agentsafe.up.railway.app/v1/scan \
  -H 'Content-Type: application/json' \
  -d '{"agent_id":"your-name","policy":"default","url":"https://example.com/SKILL.md"}'
```

Prove path first: URL-scan `https://agentsafe.up.railway.app/skill.md`.

## Detects (examples)

- remote code to shell (`curl|bash`)
- base64-decode-to-shell chains
- macOS quarantine strip (`xattr -c`) then run
- paste/bin staging hosts
- credential-sink language next to outbound calls
- crypto drain / convenience lures
- message-bus dumps of host material

## Response fields agents should read

- `report.risk_band`
- `report.decision.action` = `deny` | `review` | `allow_with_caution`
- `report.content_fingerprint` for drift checks
- `report.capability_signals`
- `next_action` (refuse / human review / allow / pay)

## Pricing

- Free daily quota always (default 20/UTC day) — free forever as a product option
- Over free: **0.05 USDC** on Base + `payment_tx` per scan
- **Packs (max-money path):** `GET /v1/packs` — Operator **29 USDC / 30d** → 500 scans/day · Fleet **99** · Watch Starter **9**
- Activate: `POST /v1/packs/activate` `{agent_id, pack_id, payment_tx}`
- Wallet: GET `/` → `pricing.receive_wallet`. ClawHub does not bill.

## Disclaimer

Heuristic only. Complements ClawHub scanners. Not legal/security/financial advice.