st-refine-plan

Pass

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local CommonJS scripts (scripts/find-strikethroo-root.cjs and scripts/validate-plan-blueprint.cjs) to resolve workspace paths and validate plan metadata.
  • [INDIRECT_PROMPT_INJECTION]: The operating procedure explicitly directs the agent to read and "execute the instructions" contained in <root>/config/hooks/PRE_PLAN.md and <root>/config/hooks/POST_PLAN.md. This creates a surface where an attacker who can modify these files (e.g., via a malicious repository or a pull request) could induce the agent to perform unauthorized actions.
  • Ingestion points: The skill reads several project-specific files including STRIKETHROO.md, PRE_PLAN.md, PLAN_TEMPLATE.md, and POST_PLAN.md from the resolved workspace root.
  • Boundary markers: The instructions do not specify any boundary markers or sanitization logic when processing the instructions within the hook files.
  • Capability inventory: The agent has the capability to read/write files and execute shell commands/scripts as part of the plan refinement process.
  • Sanitization: No sanitization or validation of the content within the hook files is performed before execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 8, 2026, 02:44 AM
Security Audit — agent-trust-hub — st-refine-plan