The Agent Skills Directory

[SAFE]: The skill instructions describe a legitimate documentation assistance process. No signs of prompt injection, data exfiltration, or malicious commands were found.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external context from user-provided info dumps and shared documents. This creates a surface for indirect prompt injection, which is handled as a core part of the co-authoring functionality. Ingestion points: Stage 1 (Context Gathering) and shared document links. Boundary markers: Not specified in instructions. Capability inventory: Use of create_file and str_replace tools. Sanitization: Not explicitly implemented.

doc-coauthoring