route-capture

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted input from file transcripts, which could contain instructions to influence agent behavior.
  • Ingestion points: The transcript content from the file specified in the argument is read and processed in SKILL.md.
  • Boundary markers: There are no explicit delimiters or specific warnings to ignore embedded instructions within the processed transcript text.
  • Capability inventory: The skill possesses file modification capabilities through Write, Edit, and limited Bash commands (mkdir, test, date).
  • Sanitization: All proposed side-effects are surfaced to the user for explicit confirmation before any writes or executions occur, serving as a primary defense.
  • [SAFE]: Tool access is appropriately restricted. The frontmatter configuration limits shell command usage to low-risk diagnostic and directory management commands.
  • [SAFE]: File operations are transparent. The skill is designed to show diffs and previews to the user for all proposed edits to files like crm/companies.md or ground-truth.md before execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 05:16 AM
Security Audit — agent-trust-hub — route-capture