The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a potential attack surface for indirect prompt injection because its primary function is to read, analyze, and modify external project code.
Ingestion points: The agent accesses untrusted data from local project files using the Read, Glob, and Grep tools defined in the frontmatter.
Boundary markers: The instructions do not establish explicit delimiters or headers to distinguish between the skill's trusted instructions and potentially malicious content found in the external files being edited.
Capability inventory: The skill possesses significant capabilities to modify the environment via the Write and Edit tools.
Sanitization: There is no evidence of content sanitization, validation, or filtering of the code ingested from the user's project before it is processed by the agent.

antislop-ui