self-healing

Warn

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes untrusted log data to generate code fixes.\n
  • Ingestion points: System and application logs from CI, Vercel, and local runtime are ingested in the 'Detect' phase in SKILL.md.\n
  • Boundary markers: No delimiters or safety instructions are defined to isolate log content from the agent's logic.\n
  • Capability inventory: The skill can modify files, commit changes to git, and execute code through build and test scripts.\n
  • Sanitization: The instructions require the 'exact' error message and stack trace, without any validation or filtering.\n- [COMMAND_EXECUTION]: The skill automatically executes shell commands to validate and apply code changes.\n
  • Evidence: Phase 4: Validate in SKILL.md executes commands including npm test, npm run build, and npx tsc --noEmit automatically after a fix is generated.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 18, 2026, 01:38 PM
Security Audit — agent-trust-hub — self-healing