The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/generate.py script executes shell commands to detect the system's language using subprocess.run(shell=True). While the commands are defined in the skill's own config.yaml, the use of shell execution is a security best-practice violation and a potential risk vector if the configuration is manipulated.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted project metadata (such as titles and descriptions from README.md) and interpolating it directly into generated AI instruction files (AGENTS.md, CLAUDE.md). A malicious project could inject instructions that influence the behavior of agents reading these documents.\n
Ingestion points: Project metadata files including README.md, pyproject.toml, and package.json processed in scripts/generate.py.\n
Boundary markers: Absent; untrusted content is inserted into Markdown templates without delimiters or instructions to ignore embedded commands.\n
Capability inventory: Local file writing for project documentation and shell command execution via language detection in scripts/generate.py.\n
Sanitization: Absent; the replace_placeholders method in scripts/generate.py performs simple string replacement without validation or escaping.

init-project