The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from session recordings, SQL query results, and person profiles. While this presents an attack surface for indirect prompt injection, the skill's capabilities are limited to data synthesis and summarization using vendor-provided tools. No high-risk actions such as local file writing or arbitrary shell command execution are present.
Ingestion points: Data is ingested via posthog:session-recording-get, posthog:persons-retrieve, and posthog:execute-sql.
Boundary markers: None explicitly defined in the prompt templates.
Capability inventory: The skill uses posthog:session-recording-summarize and internal narrative synthesis. No file-system access or subprocess execution tools are used.
Sanitization: No explicit sanitization or validation of external data is defined in the workflow.
[SAFE]: All tools (posthog:*) and data sources identified are official resources belonging to the vendor (posthog). The operations performed are consistent with the skill's stated purpose of session investigation.

investigating-replay