The Agent Skills Directory

[SAFE]: The skill provides structured workflows for investigating tool error rates, latency, and failure reasons using PostHog's internal HogQL and specialized diagnostic tools.
[SAFE]: All external links point to official PostHog application domains (app.posthog.com), which are trusted for this vendor resource.
[SAFE]: The skill encourages best practices by instructing the agent to cast database properties (e.g., toBool, toFloat) and to check the governed metrics catalog before running manual queries.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by ingesting and processing external telemetry data (e.g., error messages or tool names captured in the events table).
Ingestion points: Data retrieved from the shared events table through tools like posthog:execute-sql and posthog:query-mcp-tool-failures.
Boundary markers: None identified; there are no specific instructions to wrap tool outputs in delimiters or ignore embedded instructions within retrieved data.
Capability inventory: The skill's primary capabilities are data aggregation, performance reporting, and generating dashboard links.
Sanitization: The provided SQL templates enforce type casting for numeric and boolean fields, but the skill does not explicitly sanitize free-text fields like $mcp_error_message.

exploring-mcp-tool-quality