lodestar-fix
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external files produced by other tools (INDEX.md, findings.md, and action-item markdown files) to determine which code modifications to apply. This represents an attack surface where malicious data in those files could influence agent behavior.
- Ingestion points:
INDEX.md,findings.md, and individual action-item markdown files in the audit output directory. - Boundary markers: The skill instructions emphasize that the action item is the "contract" and define "Resident safety" rules to restrict edits to specific file lists.
- Capability inventory: The skill can execute
gitcommands (add, commit, restore), run local node scripts, and execute project-defined validation commands (typecheck,lint,test). - Sanitization: The
action-state.mjsscript includesassertCommitLineto validate that generated git commit subjects and trailers do not contain newlines or control characters. - [DYNAMIC_EXECUTION]: The
setup-modules.mjsscript uses dynamicimport()to load a shared runtime module (runtime.mjs) from a relative path within the vendor's skill ecosystem. - Evidence:
setup-modules.mjscontainsawait import(url)where the URL points to../../lodestar-setup/scripts/runtime.mjsrelative to the current file. - [COMMAND_EXECUTION]: The skill executes various shell commands to perform its primary function of fixing code and verifying changes.
- Evidence: The skill runs
git status,git add,git commit, andgit restore. It also executes commands configured in the project environment for linting, typechecking, and testing as defined in.agents/lodestar/context.md.
Audit Metadata