The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from Notion workspaces, which serves as an untrusted data ingestion point. If retrieved pages or databases contain malicious instructions, they could influence the agent's behavior during subsequent processing steps.
Ingestion points: Data is ingested via notion.search(), notion.databases.query(), and notion.blocks.children.list() as seen in SKILL.md and references/implementation.md.
Boundary markers: The provided implementation does not use explicit delimiters or instructions to ignore commands embedded within the retrieved Notion content.
Capability inventory: The skill operates in an environment with Bash, Write, and Edit capabilities, which could be leveraged if the agent is successfully manipulated by injected content.
Sanitization: Content extraction focuses on plain text retrieval but lacks specific sanitization or filtering to detect or neutralize prompt injection patterns.

notion-search-retrieve