plan-happy-path

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates a surface for indirect prompt injection by incorporating raw user-provided implementation requests into prompts for sub-agents without isolation.\n
  • Ingestion points: The user's request is captured verbatim in the 'Input' section and passed to the planner child in Step 1 and the verifier child in Step 2 of the Workflow.\n
  • Boundary markers: There are no boundary markers or delimiters (such as XML tags) used to separate the untrusted user request from the skill's instructions during interpolation.\n
  • Capability inventory: The skill uses the rlm(...) tool to spawn sub-agents and includes file system capabilities to create directories and write Markdown files in the .scratch/plans/ path.\n
  • Sanitization: The instructions do not include any sanitization, validation, or escaping logic for the incoming user request before it is used to drive agent behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 11:54 AM
Security Audit — agent-trust-hub — plan-happy-path