skill-safety-audit
Pass
Audited by Gen Agent Trust Hub on Jul 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script (
scripts/source_ingestion_guardrail.py) as part of its safety audit workflow. This is a functional requirement for the skill's purpose of detecting unauthorized content and occurs within the local repository environment. - [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because its primary function is to read and analyze potentially untrusted files (other skills) which could contain adversarial instructions. The skill provides specific workflow guidelines to mitigate this, such as reading instructions in full before trusting examples and ignoring "rubber-stamping" tendencies.
- Ingestion points: Reads changed skill files and bundled resources from the filesystem and git diff.
- Boundary markers: The workflow explicitly instructs the auditor to inspect instructions before trusting scripts or examples.
- Capability inventory: The skill can read/search the filesystem and execute local Python scripts.
- Sanitization: No programmatic sanitization is mentioned; security relies on the LLM's adherence to the audit workflow.
Audit Metadata