The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the posthog:execute-sql tool to run HogQL queries against the project's event database. This is the primary intended function of the skill to perform cost aggregations and trend analysis.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from event properties (such as $ai_model, $ai_provider, and custom user-defined properties) when performing breakdowns and filters.
Ingestion points: Data enters the agent context via the events table in PostHog, specifically from $ai_generation and $ai_embedding events in files like SKILL.md and references/breakdown-patterns.md.
Boundary markers: Not explicitly defined for property values, as they are processed as data fields within SQL aggregations.
Capability inventory: The skill has the ability to execute SQL queries, create insights, dashboards, and alerts through the PostHog API.
Sanitization: The skill encourages using the posthog:read-data-schema tool to discover valid property names rather than guessing, and uses a specialized tool (posthog:generate-app-url) for safe URL construction to avoid region/project mismatches.

exploring-llm-costs