skills/bryantleft/bruhs-skills/bruhs/Gen Agent Trust Hub

bruhs

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it handles external data at runtime.
  • Ingestion points: The skill ingests untrusted data in commands/cook.md (via Linear ticket descriptions fetched via MCP) and commands/peep.md (via PR review comments fetched via the GitHub CLI).
  • Boundary markers: The prompts defined in commands/peep.md and the workflows in commands/cook.md do not utilize explicit delimiters (e.g., XML tags) or instructions for the agent to ignore any embedded AI instructions within the fetched text.
  • Capability inventory: The skill possesses significant capabilities, including arbitrary command execution (Bash), file modification (Edit), and interaction with authenticated CLI tools (gh) and MCP servers (Linear).
  • Sanitization: There is no evidence of filtering or sanitization performed on the external content before it is processed by subagents.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 01:29 AM
Security Audit — agent-trust-hub — bruhs