The Agent Skills Directory

[SAFE]: The skill is designed for diagnostics and root-cause analysis. It operates with a 'read-only' design principle, explicitly instructing the agent not to change code or perform destructive operations.
[COMMAND_EXECUTION]: The skill uses Bash, Grep, Read, and Glob to analyze the local repository and connect trace data to the source code. These are standard operations for a developer-oriented diagnostic skill.
[EXTERNAL_DOWNLOADS]: The skill fetches data from Opik (comet.com) via a hosted MCP or the Python SDK (opik). Since the skill is authored by comet-ml, these network operations target their own official infrastructure for their intended purpose.
[CREDENTIALS_UNSAFE]: The skill mentions checking for OPIK_API_KEY or ~/.opik.config. It does not hardcode these secrets, but rather follows standard practice by checking environment variables or local configuration files managed by the user.

opik-explain