The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source files from a project directory to generate markdown documentation, creating a surface for indirect prompt injection via malicious filenames or path structures. \n
Ingestion points: Project source files scanned by scanProject in scripts/wiki.js and scan_project in scripts/wiki.py. \n
Boundary markers: Filenames are enclosed in markdown backticks within the generated registration tables, providing basic structural delimitation. \n
Capability inventory: The skill has file-writing capabilities to create and update markdown files in the docs/project_wiki/ directory. \n
Sanitization: No explicit escaping or sanitization of path strings was found before interpolation into the markdown templates. \n- [COMMAND_EXECUTION]: The skill's test suite executes the bundled Python and Node.js scripts using subprocess calls for functional verification. \n
Evidence: tests/conftest.py utilizes subprocess.run to invoke scripts/wiki.py and scripts/wiki.js against temporary fixture projects.

project-wiki