The Agent Skills Directory

[SAFE]: No malicious patterns or unauthorized behaviors were detected. The skill is consistent with its stated purpose of documenting analytics queries for the PostHog platform.
[DATA_EXFILTRATION]: The skill describes how to query the document_embeddings table to fetch product observations using the posthog:execute-sql tool. This access is inherent to the skill's purpose and is constrained by the platform's team-level permissions.
[PROMPT_INJECTION]: The processing of natural-language signal descriptions presents a surface for indirect prompt injection.
Ingestion points: Signal descriptions are retrieved from the content column of the document_embeddings table via the posthog:execute-sql tool.
Boundary markers: The provided SQL templates do not define explicit delimiters to separate untrusted signal data from agent instructions.
Capability inventory: The skill utilizes the posthog:execute-sql capability for data retrieval.
Sanitization: No sanitization or validation logic is specified for the data ingested from the database.

signals