The Agent Skills Directory

[SAFE]: The skill is a legitimate tool for the Coval evaluation platform. It performs expected tasks such as plan validation, launching simulations via a CLI, and fetching results for auditing.
[COMMAND_EXECUTION]: The skill uses the coval CLI and local Python scripts (scripts/plan.py, scripts/fetch_run.py) to perform its duties. These are standard operations for a developer-oriented tool.
[EXTERNAL_DOWNLOADS]: The scripts/fetch_run.py script connects to https://api.coval.dev/v1 to retrieve conversation transcripts and metric results. This is a vendor-owned API and is necessary for the skill's functionality. The script uses standard libraries and implements pagination and identity checks to ensure data integrity.
[CREDENTIALS_SAFE]: The skill correctly handles sensitive information by instructing users to provide the COVAL_API_KEY via environment variables rather than hardcoding it. The fetch_run.py script specifically writes output files with restrictive permissions (0o600) to prevent unauthorized access to local snapshots.

quick-eval