canary-watch
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill monitors external URLs and processes their content, console logs, and network responses to verify deployment health. This design provides a surface where a malicious or compromised website could embed instructions targeting the AI agent.
- Ingestion points: Fetched HTML content, console error logs, and network response bodies from external URLs specified in the
/canary-watchcommand. - Boundary markers: None; the instructions do not include specific delimiters or directives to ignore instructions that might be contained within the monitored web content.
- Capability inventory: The skill performs network requests (HTTP and SSE streams) and writes execution results to a local log file at
~/.claude/canary-watch.log. - Sanitization: No sanitization, filtering, or validation steps for the external data are described before the agent processes the information.
Audit Metadata