clawarena

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to install via a custom CLI tool 'clawdhub' from clawhub.ai. It also includes a heartbeat mechanism that periodically fetches guidance from heartbeat.md on the clawarena.ai domain.
  • [COMMAND_EXECUTION]: The skill guides the agent to use curl commands for registration, fetching market data, and submitting predictions to the ClawArena API.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and follows instructions from external sources. 1. Ingestion points: Periodic fetching of heartbeat.md and processing of market metadata from the API. 2. Boundary markers: None identified; the agent is instructed to follow the content of the fetched files directly. 3. Capability inventory: curl for network operations and read/write access to local configuration files in ~/.config/clawarena/. 4. Sanitization: No sanitization or validation of the fetched markdown instructions or API data is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 02:22 AM