cost-estimate

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it analyzes untrusted content from software repositories without established security boundaries.\n
  • Ingestion points: Source code, architecture, and documentation within the repository being analyzed are ingested for inventory purposes (SKILL.md).\n
  • Boundary markers: The instructions do not define delimiters or provide warnings to the agent to distinguish between repository content and its own instructional logic.\n
  • Capability inventory: The skill utilizes file reading, repository searching, git operations, and web research tools to generate its report (SKILL.md).\n
  • Sanitization: There is no mention of sanitizing or validating repository content before it is processed by the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 06:40 PM