The Agent Skills Directory

[SAFE]: The skill instructions are entirely focused on collaborative writing and follow a transparent, user-driven workflow. No evidence of prompt injection, privilege escalation, or persistence mechanisms was found. It uses standard agent tools like 'create_file' and 'str_replace' for their intended purposes.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it is designed to ingest and process external context from user dumps and shared documents. (1) Ingestion points: User info-dumps and shared documents/links from integrations. (2) Boundary markers: The skill does not provide specific instructions to delimit untrusted context. (3) Capability inventory: The agent has access to 'create_file' and 'str_replace' tools to modify the workspace. (4) Sanitization: The skill lacks explicit sanitization logic, relying on the underlying platform's security guardrails. This risk is inherent to the primary function of the skill.

doc-coauthoring