The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides tools and instructions for fetching and reading external content (URLs, code files) for the purpose of development and UI review. This capability creates an attack surface for indirect prompt injection, where malicious instructions embedded in the external data could attempt to influence the agent's behavior.
Ingestion points: The skill allows the use of the WebFetch tool to retrieve external content in SKILL.md, ui-review-tahoe/SKILL.md, and macos-tahoe-apis/SKILL.md. Additionally, it reads project files using Read, Glob, and Grep.
Boundary markers: The skill instructions do not explicitly define boundary markers or include warnings for the agent to ignore instructions embedded within the retrieved external content.
Capability inventory: The skill has access to Read, Glob, Grep, WebFetch, Write, and AskUserQuestion tools.
Sanitization: There is no mention of sanitization, validation, or filtering of the content ingested from external sources before it is processed by the agent.

macos-development