privacy-safe-skill-extraction
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process internal documentation, runbooks, and workflows, creating a surface for indirect prompt injection if the source material contains adversarial instructions.\n
- Ingestion points: Internal materials provided for sanitization as described in SKILL.md.\n
- Boundary markers: Instructions in SKILL.md require clear separation between source and output, though technical delimiters for the agent's input stream are not defined.\n
- Capability inventory: No dangerous capabilities such as shell commands, network requests, or file system modifications are used in SKILL.md or evals/evals.json.\n
- Sanitization: The skill's primary function in SKILL.md is the identification and removal of sensitive data, including credentials, PII, and internal infrastructure identifiers.
Audit Metadata