The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data snippets retrieved through the brain_search tool, which presents a potential surface for indirect prompt injection from the contents of the knowledge base.
Ingestion points: Snippets returned by the mcp__governed-brain__brain_search tool as described in the search workflow in SKILL.md.
Boundary markers: While explicit delimiters for data snippets are not defined in the instructions, the skill enforces strict grounding by requiring qmd:// citations for every claim and synthesis based only on returned snippets.
Capability inventory: The skill utilizes the mcp__governed-brain__brain_search tool. Prerequisite documentation in SKILL.md and references/runtime-contract.md mentions a requirement for a local qmd binary on the system path for local mode operations.
Sanitization: The instructions provide a strict multi-step retrieval process and mandate that the agent must refuse to answer if no results are found, effectively preventing the use of unverified general knowledge.

brain