The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use curl for interacting with the Warden Service API. It follows security best practices by utilizing environment variables (WARDEN_PAT) for authentication and --data-urlencode for query parameters to prevent shell injection and credential leakage in command history.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where it ingests and summarizes external JSON data from the Warden Service API.
Ingestion points: Data is fetched from the Warden Service API via routes like /api/v1/findings and /api/v1/runs as described in SKILL.md and references/read-api.md.
Boundary markers: The instructions lack explicit boundary markers for the LLM during the summary phase, though they constrain the agent to summarize only "relevant JSON fields."
Capability inventory: The skill uses curl for network operations and provides summaries to the user.
Sanitization: Input parameters are sanitized using URL encoding; however, the content of the API response itself is processed directly for summarization.
[CREDENTIALS_UNSAFE]: The skill explicitly includes behaviors and constraints to protect sensitive information. It mandates that the agent must not request, print, or embed the WARDEN_PAT personal access token in plain text or logs, directing users to safe configuration methods instead.

warden-service