security-threat-model

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection as it processes untrusted content from external repositories to generate security analysis. [1] Ingestion points: The skill reads files and metadata from user-specified repository directories (references/prompt-template.md). [2] Boundary markers: Markdown formatting and headers are used to delimit content, but explicit 'ignore instructions' blocks are not implemented. [3] Capability inventory: The skill has file system read access for analysis, file writing to save reports, and shell execution for search tools like ripgrep. [4] Sanitization: No explicit input validation or filtering of codebase content is mentioned.
  • [COMMAND_EXECUTION]: The skill instructions (references/prompt-template.md) recommend using the ripgrep (rg) command-line utility for searching the repository. This is a legitimate use of tooling for the skill's primary purpose of code analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 02:16 PM
Security Audit — agent-trust-hub — security-threat-model