The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external trace and diagnostic issue data from the Opik platform, which represents a potential surface for indirect prompt injection if the ingested data contains malicious instructions.
Ingestion points: agent_insights_issue entities and trace search results (e.g., error messages, span content) via MCP or the Python SDK.
Boundary markers: The instructions do not define specific delimiters or warnings to treat ingested trace data as untrusted.
Capability inventory: The skill has access to Bash, Read, Grep, and Glob tools.
Sanitization: No explicit sanitization of trace data is mentioned before the information is presented to the user or handed off to other skills.
[COMMAND_EXECUTION]: The skill uses the Bash tool to verify the environment, specifically checking for the presence of ~/.opik.config or the OPIK_API_KEY environment variable. The instructions focus on read-only diagnostics and do not involve execution of arbitrary or attacker-supplied commands.
[CREDENTIALS_UNSAFE]: The skill checks for configuration at standard paths (~/.opik.config) and in environment variables to authenticate with the Opik service. This behavior is consistent with standard service integration and does not involve hardcoding secrets or exfiltrating credentials.

opik-diagnose