brainstorm-beagle

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted external data during its discovery phase.
  • Ingestion points: The skill reads existing .beagle/concepts/<slug>/brief.md files, project documentation, general project files, and git history to understand context (defined in SKILL.md, Steps 1 and 2).
  • Boundary markers: The instructions lack specific delimiters or negative constraints to prevent the agent from following instructions potentially embedded in these source files.
  • Capability inventory: The skill can write new specification files to the disk and execute git commit commands (defined in SKILL.md, Step 9).
  • Sanitization: No explicit sanitization or filtering of the content read from files is described before it is integrated into the brainstorming dialogue or the final spec document.
  • [COMMAND_EXECUTION]: The skill is designed to interact with the file system and git. It writes to the .beagle/ directory and executes git commits (e.g., git commit -m "docs: add <slug> project spec"). While these are intended functional behaviors, they constitute the capability component of the skill's vulnerability profile.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 04:41 PM
Security Audit — agent-trust-hub — brainstorm-beagle