The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is intended to evaluate external technical documents, RFCs, and architecture designs provided by the user or fetched from the web. It lacks specific instructions to handle these inputs as untrusted, creating a surface for indirect prompt injection attacks where instructions hidden in the data could override the agent's behavior.
Ingestion points: The skill utilizes Read, Glob, Grep, WebSearch, and WebFetch tools to ingest content from the local filesystem and the internet (SKILL.md).
Boundary markers: The instructions do not define delimiters or provide 'ignore embedded instructions' warnings for the data it processes.
Capability inventory: The skill possesses read-only access to files and web content but does not have shell execution or file-write capabilities.
Sanitization: There is no evidence of sanitization or filtering logic applied to content before it is interpolated into the agent's reasoning process.

engineering-expert