woo-sailor

Pass

Audited by Gen Agent Trust Hub on Mar 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill directly interpolates user-provided arguments ($0, $1, $ARGUMENTS) into the prompt structure within XML tags. An attacker could provide crafted input containing closing tags (e.g., ) to break out of the intended template and inject arbitrary instructions to the agent.\n- [PROMPT_INJECTION]: The skill identifies and processes project configuration and rule files as untrusted data, creating an indirect prompt injection surface.\n
  • Ingestion points: Files matching patterns such as **/SKILL.md, **/CLAUDE.md, and **/AGENT.md are targeted for processing.\n
  • Boundary markers: No delimiters or 'ignore embedded instructions' warnings are used when the sub-agent is instructed to process these files.\n
  • Capability inventory: The delegated sub-agent (process-siren) has the capability to perform 'in-place' file system modifications.\n
  • Sanitization: The skill does not perform validation or filtering of the content within the target files before they are processed for optimization.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 29, 2026, 08:42 AM