The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted user data, creating a potential surface for indirect prompt injection attacks.
Ingestion points: The "Mode: critique" workflow in SKILL.md explicitly takes a user-provided draft for scoring and rewriting.
Boundary markers: The instructions do not specify any delimiters or special markers to isolate the user's draft from the agent's internal logic.
Capability inventory: The agent has access to Read, Write, and Grep tools, which could potentially be misused if the agent is manipulated by instructions hidden within a draft.
Sanitization: There are no instructions for sanitizing or validating the input text before processing it.

founder-brand