patcher

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from the project being analyzed.
  • Ingestion points: The agent is instructed in SKILL.md to gather vulnerability inputs from package manifests (package.json, pyproject.toml, etc.), lockfiles, Dockerfiles, and CI/CD configuration files.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when processing content from these external files.
  • Capability inventory: The skill enables the agent to perform file-write operations (for patching) and execute shell commands (for auditing and verification) via scripts/check-patcher.mjs.
  • Sanitization: The instructions do not require sanitization or validation of content found in the ingested files before it is processed by the LLM.
  • Remediation: To mitigate this risk, external content should be wrapped in delimiters with an explicit 'ignore instructions' warning, and the skill should prefer structured data parsing with strict schema validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 01:34 AM
Security Audit — agent-trust-hub — patcher