The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to fetch and analyze "signals"—natural-language descriptions of product events such as error messages or experiment results. Because this content originates from external sources (e.g., error tracking payloads), it may contain malicious instructions designed to manipulate the agent's behavior when the content is processed or summarized.
Ingestion points: The content and metadata columns of the document_embeddings table referenced in SKILL.md contain data from external product sources.
Boundary markers: The provided HogQL query examples do not include instructions for the agent to use delimiters or ignore potentially malicious instructions embedded within the signal content.
Capability inventory: The skill utilizes the posthog:execute-sql tool to perform database operations.
Sanitization: No sanitization, validation, or escaping of the content field is performed before the agent reads and processes the text.

signals