The Agent Skills Directory

[SAFE]: The skill implements significant security controls for handling sensitive data. It uses atomic file operations via os.open() with 0o600 permissions and enforces a 0o700 mode on its storage directory to prevent unauthorized access. The RestClient implementation in scripts/_providers_core.py further secures data by requiring HTTPS for authenticated requests and redacting credentials from all error outputs.
[COMMAND_EXECUTION]: The skill uses shell commands for repository identification and authentication verification. In scripts/_list_providers.py, it executes git remote get-url origin to resolve provider URLs. In scripts/_providers_core.py, the verify_system_auth function executes a verification command (e.g., gh auth status) provided by local provider extensions to confirm tool status.
[REMOTE_CODE_EXECUTION]: The skill uses a plugin-based architecture for provider discovery. In scripts/_list_providers.py, the _load_provider_module function uses importlib.util to load Python modules matching *_provider.py from the PYTHONPATH. This is a core architectural feature used to register credential providers.
[DATA_EXFILTRATION]: While the skill communicates with external APIs, it is configured to do so safely. The RestClient redacts sensitive tokens and passwords from response bodies and ensures that credentials never leak into the agent's logs or tracebacks. The skill's design explicitly prevents the AI agent from accessing the raw credential files.
[PROMPT_INJECTION]: The skill processes data from project configuration files, representing a potential surface for indirect prompt injection. 1. Ingestion points: scripts/_cred_configure.py parses pom.xml files to auto-derive configuration properties. 2. Boundary markers: The script validates user-supplied values against project properties and issues warnings for mismatches. 3. Capability inventory: The skill includes file system access, command execution, and module loading capabilities. 4. Sanitization: Project names are sanitized via regex to prevent path traversal.

manage-providers