ezbot-agent-id-feed
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for external project data (names, summaries) which may contain malicious instructions meant to influence the AI agent.\n
- Ingestion points: Remote data is fetched in
scripts/feed_client.pyfrom a JSON manifest and catalog.\n - Boundary markers: Data is wrapped in markdown blockquotes in the
renderfunction, andSKILL.mdcontains explicit instructions for the agent to treat this content as untrusted data.\n - Capability inventory: The skill can execute local Python scripts and read/write to a local state file in the user's home directory.\n
- Sanitization: The
display_textfunction inscripts/feed_client.pyapplies NFKC normalization and regex-based redaction to block common prompt injection phrases and sensitive data markers (e.g., 'ignore previous instructions', 'private key').\n- [EXTERNAL_DOWNLOADS]: Fetches signed manifest and catalog files from a Cloudflare Workers URL (jaxonly.workers.dev).\n- [COMMAND_EXECUTION]: Executes thefeed_client.pyscript via the Python interpreter to manage data subscriptions and filtering, which is the primary functionality of the skill.
Audit Metadata