evidence-recap

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious patterns such as credential exfiltration, persistence mechanisms, or unauthorized command execution were detected. The skill is entirely instructional and operates within a text-based dialogue summarization context.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of programming dialogue history, creating a surface for indirect prompt injection.
  • Ingestion points: Dialogue segments and project records processed in SKILL.md.
  • Boundary markers: The instructions do not define specific delimiters (e.g., XML tags or triple backticks) to isolate the dialogue input.
  • Capability inventory: The skill has no capabilities to execute shell commands, write files, or perform network requests.
  • Sanitization: The skill includes a robust privacy check step that instructs the AI to generalize or replace project codes, tokens, keys, emails, and internal paths before outputting results.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 12:52 PM
Security Audit — agent-trust-hub — evidence-recap