lodestar-implement
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill identifies and runs shell commands (e.g., test, lint, typecheck) found in the repository's configuration files like package.json. This is a standard but sensitive operation for a development tool.
- [INDIRECT_PROMPT_INJECTION]: The skill processes project-level files and plan documentation which could contain instructions intended to influence the agent's behavior.
- Ingestion points: Plan files and .agents/lodestar/context.md are read and parsed by the plan-state.mjs script and locate.md reference.
- Boundary markers: There are no explicit delimiters or warnings to the model to ignore instructions embedded within the data it processes.
- Capability inventory: The skill can modify files, execute git operations, and run discovered project scripts.
- Sanitization: The skill does not perform verification or sanitization of the command strings discovered from project metadata before execution.
- [DYNAMIC_EXECUTION]: The setup-modules.mjs script uses dynamic imports to load shared functionality from a relative sibling directory (../../lodestar-setup/scripts/).
- [PROMPT_INJECTION]: Instructions in discover-context.md constrain the agent's communication with the user, specifically directing it not to recommend certain actions or to limit mentions of specific tools.
Audit Metadata