The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Node.js scripts (scan-codebase.cjs, generate-intel.js, detect-stale-intel.cjs) that invoke execSync to interact with the local Git binary. These operations are limited to metadata retrieval (e.g., git rev-parse, git ls-files, git diff) and standard repository management (git add, git commit).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase data to generate documentation, presenting an inherent injection surface.
Ingestion points: Files are read directly from the project directory using fs.readFileSync within scan-codebase.cjs and summarized by subagents as described in codebase-mapper-instructions.md.
Boundary markers: The skill uses structured Markdown templates and JSON schemas for output, though it does not implement explicit "ignore embedded instructions" delimiters for the content it analyzes.
Capability inventory: The skill possesses capabilities for shell command execution (Git), file system writes, and subagent orchestration.
Sanitization: scan-codebase.cjs employs regex-based comment stripping to reduce the risk of processing malicious instructions embedded in source code comments.
[DATA_EXPOSURE]: The skill templates (e.g., INTEGRATIONS.md) correctly instruct the agent to identify environment variable names rather than hardcoding actual secrets or API keys. It also properly recognizes that .env files should be ignored or used only for reference.

kata-map-codebase