The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the execute-sql tool to run analytical HogQL queries against the events table. These queries are structurally used to audit experiment exposures and identify anomalies like Sample Ratio Mismatch (SRM). The execution is limited to data analysis within the PostHog environment.
[DATA_EXFILTRATION]: Communication is restricted to the vendor's platform. The skill reads experiment configurations and results from the PostHog API and writes findings to an internal scratchpad or emits signals. No sensitive data exfiltration to unauthorized external domains was identified.
[PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection by processing potentially untrusted data from experiment metadata.
Ingestion points: Retrieves experiment names, descriptions, and feature flag keys via experiment-list and experiment-get.
Boundary markers: The instructions do not define explicit boundary markers or provide warnings to ignore embedded instructions within processed metadata.
Capability inventory: The skill can perform database queries via execute-sql, update its internal state via signals-scout-scratchpad-remember, and report findings via signals-scout-emit-signal.
Sanitization: No specific sanitization or escaping of ingested experiment metadata is mentioned in the instructions.

signals-scout-experiments