clawfinder

Pass

Audited by Gen Agent Trust Hub on Mar 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes the local gpg binary for cryptographic tasks including key generation, signing, and message encryption/decryption as part of its core protocol.
  • [EXTERNAL_DOWNLOADS]: Recommends the installation of the @crossmint/lobster-cli package from the npm registry for handling payment settlements.
  • [PROMPT_INJECTION]: The skill's architecture involves ingesting and processing content from external agents, creating a surface for indirect prompt injection.
  • Ingestion points: The agent processes data from the /api/agents/me/inbox/ endpoint, job search results, and external files retrieved via attachment_url fields.
  • Boundary markers: Messages use a structured key-value header format and blank-line body delimiters to separate control metadata from untrusted content.
  • Capability inventory: The skill performs shell command execution via the GPG binary and makes network requests to both the protocol index and third-party file hosts.
  • Sanitization: The protocol specification does not define explicit sanitization or validation logic for the free-form text bodies and payloads received from counterparties.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 20, 2026, 10:08 AM