The Agent Skills Directory

[SAFE]: The skill operates entirely as a local diagnostic utility. It relies exclusively on the Python standard library and does not perform network operations, external downloads, or data exfiltration.
[COMMAND_EXECUTION]: The testing utility scripts/run_fixtures.py uses subprocess.run to execute the companion triage script against a matrix of local fixtures. This execution is limited to internal testing, does not invoke a shell (shell=True is not used), and relies on arguments derived from local configuration files.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface for analyzing external data such as OAuth token records and connector event logs.
Ingestion points: Data is ingested through the --token-file and --log command-line arguments in scripts/oauth_triage.py.
Boundary markers: Documentation explicitly warns users to redact secrets before use, and the tool is presented as a diagnostic utility separate from operational production code.
Capability inventory: The skill's capabilities are restricted to reading local files and executing internal test scripts; it has no ability to write to arbitrary filesystem locations or communicate over the network.
Sanitization: The tool includes a dedicated redact() function in scripts/oauth_triage.py that filters common OAuth secret keys (e.g., access_token, refresh_token) from output to protect user credentials.

mcp-oauth-triage