project-structure
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to read source files from external repositories and summarize them for an LLM. This creates a vulnerability where malicious instructions hidden in code comments or symbol names within the target repository could be processed by the agent.
- Ingestion points:
map.tswalks and reads all source files within a user-definedrepoRootdirectory. - Boundary markers: While the output uses file headers, it lacks explicit delimiters or instructions to the model to ignore embedded instructions in the mapped content.
- Capability inventory: The skill performs file reads across the target repository and writes output to the local filesystem using
node:fs. - Sanitization: The script truncates long signatures and doc-comments but does not sanitize or filter the content for prompt injection patterns.
- [DYNAMIC_EXECUTION]: The script uses dynamic loading to resolve the
typescriptpackage from the target repository's environment. - Evidence:
map.tsusescreateRequirewith a path derived from the target repository to load thetypescriptcompiler. - [COMMAND_EXECUTION]: The skill requires the execution of Node.js scripts to perform its analysis and testing.
- Evidence:
SKILL.mdinstructs the agent to runnode map.tsandselftest.mjsusesexecFileSyncto run the mapper.
Audit Metadata