The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data from local files and third-party platforms (such as GitHub, Figma, and Notion) into the agent's context, creating a potential vector for malicious instructions.\n
Ingestion points: The documentation in SKILL.md describes workflows for uploading files via filePath and creating link attachments via external url parameters.\n
Boundary markers: The provided code snippets and instructions do not include specific boundary markers or "ignore embedded instructions" warnings to isolate data from agent instructions.\n
Capability inventory: The skill allows access to high-capability tools such as Bash, Read, Grep, and Glob, and utilizes network operations via fetch to interact with external services.\n
Sanitization: The skill recommends metadata-level sanitization (e.g., stripping EXIF data, validating content-type), but does not address the risk of natural language instructions embedded within the data content or URL previews themselves.

Linear Attachments