The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it is designed to ingest and store untrusted data from external platforms.
Ingestion points: Data enters the system as snippet content in the evidence_item table via the ansor / create_record tool call, pulling from sources like Slack, Linear, and HubSpot (SKILL.md).
Boundary markers: The skill does not provide instructions or patterns for using delimiters or "ignore embedded instructions" warnings when processing or retrieving these data snippets.
Capability inventory: The skill utilizes the ansor MCP server to perform database queries and record creation/updates within the product_os database (SKILL.md).
Sanitization: There is no documentation or implementation of sanitization, escaping, or validation for content received from external integrations before it is stored or displayed to other agents.

ansor-memory