best-practices-security
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a security guidance framework. It instructs the agent to utilize established industry standards (OWASP ASVS, Node.js hardening guides) to perform authorized security evaluations.
- [EXTERNAL_DOWNLOADS]: The skill references several external resources from trusted organizations (OWASP, GitHub repos of known developers). These are documented as canonical baselines for context and do not involve remote code execution.
- [COMMAND_EXECUTION]: The skill mentions
/code-runnerand/hacktools for generating 'proof probes'. However, it explicitly defines 'Bounded Proof-Probe Rules' that require probes to be non-destructive, non-persistent, and executed within bounded Docker containers rather than the host system. - [PROMPT_INJECTION]: The skill includes instructions to separate memory/RAG content by trust level and treats LLM output as untrusted before reaching sensitive sinks (shell, SQL, filesystem), which is a defensive measure against injection attacks.
- [REMOTE_CODE_EXECUTION]: The
fixtures/agentic_eval.jsonfile contains a test case that executes a local Python script (validate_skill.py) to verify the skill's contract. This is a standard development testing pattern and does not involve remote code execution from untrusted sources.
Audit Metadata