The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/validate.py and its test suite scripts/test_validate.py use the subprocess and sys.executable modules. A detailed audit of this code confirms it is used strictly for a deterministic validation gate (running JSON schema and algorithmic checks) on the skill's own output files. It does not ingest untrusted user input into shell commands or execute remote code.
[DATA_EXFILTRATION]: No network-capable tools (like curl or wget) or sensitive file access patterns were detected. The workflow relies on local references and schemas provided within the skill package.
[PROMPT_INJECTION]: The skill includes a specific requirement for an external-content-sanitizer skill intended to neutralize potential prompt injections in external content, demonstrating a proactive security posture.
[CREDENTIALS_UNSAFE]: The manifest-schema.md reference provides examples of credential management (e.g., Personal Access Tokens, API Keys) but correctly instructs that these fields must remain null during the discovery phase, following safe secret management practices.

project-document-discovery