The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external data, including chat messages from other participants, reports from subagents, and the current state of PRs or CI runs. This creates a potential surface where malicious instructions embedded in those external sources could attempt to influence the agent's behavior.
Ingestion points: SKILL.md directs the agent to gather data from chat history, subagent reports, and PR/CI statuses.
Boundary markers: The skill does not employ explicit delimiters for external content but includes a primary directive to "Keep project state read-only."
Capability inventory: The agent's capabilities are restricted to summarization; it is explicitly prohibited from changing code, starting services, or posting external content.
Sanitization: No explicit sanitization or filtering of the ingested data is defined in the instructions.
[EXTERNAL_DOWNLOADS]: The skill metadata in MIRROR.md references the official GitHub repository for Cursor plugins as its upstream source.

cursor-ketchup