threat-model
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill involves analyzing external codebases and documentation, which constitutes a vulnerability surface for indirect instructions.
- Ingestion points: SKILL.md specifies the analysis of repository content, policies, and imported documents.
- Boundary markers: The Procedure section explicitly instructs the agent to treat repository content and imported documents as untrusted data.
- Capability inventory: None. The skill does not contain executable scripts or request dangerous tools in its frontmatter.
- Sanitization: The instructions explicitly command the agent to never execute target code or fetch embedded URLs during the analysis process.
- [SAFE]: The skill is a legitimate instructional guide for security auditing. No malicious patterns such as direct prompt injection, data exfiltration, or obfuscation were identified.
Audit Metadata