The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands including find, grep, and ls to traverse the workspace and extract information from project files (e.g., scanning .csproj and package.json).
[COMMAND_EXECUTION]: It uses node -e to run inline JavaScript that loads and executes local project files such as ./.claude/hooks/lib/project-config-schema.cjs and ./.claude/hooks/tests/test-all-hooks.cjs.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface (Category 8) because it ingests data from arbitrary project files to populate the configuration.
Ingestion points: Reads docs/project-config.json, package.json, .csproj files, and source code contents across the src/ directory.
Boundary markers: No specific delimiters or "ignore embedded instructions" markers are applied to the data being scanned and processed.
Capability inventory: The skill can execute shell commands, run local Node scripts, and write to the file system (updating project-config.json and creating CLAUDE.md).
Sanitization: It uses a local schema validator (project-config-schema.cjs) to ensure the resulting JSON matches the expected structure, which provides a layer of structural validation but does not prevent malicious content in string values.

project-config