plan-agent

Pass

Audited by Gen Agent Trust Hub on Jun 22, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use the ls command to verify the existence of a codebase map in the session's handoff directory.
  • [DATA_EXPOSURE]: The agent reads local codebase files and conversation context to generate planning documents. Output is written to local directories such as thoughts/shared/plans/ and thoughts/handoffs/.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from the conversation context and external codebase files which could potentially contain malicious instructions.
  • Ingestion points: The skill ingests 'Conversation context' and files identified by 'scout' agents (referenced in Step 2 and Step 3).
  • Boundary markers: No explicit boundary markers or isolation instructions are provided within the skill to distinguish between system instructions and processed data.
  • Capability inventory: The skill has the capability to spawn sub-agents (scout), read files, write files, and execute basic shell commands (ls).
  • Sanitization: No sanitization or validation logic is specified for the data retrieved from the codebase or conversation history.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 22, 2026, 07:46 AM
Security Audit — agent-trust-hub — plan-agent