doc-coauthoring
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it actively encourages and implements the ingestion of data from external, untrusted sources.
- Ingestion points: In
SKILL.md, Stage 1 ("Context Gathering") explicitly instructs the agent to read content from team channels, chat threads, and shared documents (e.g., via Slack, Teams, Google Drive, or SharePoint integrations). - Boundary markers: There are no instructions provided to wrap the ingested external content in delimiters or to include system-level warnings for the agent to ignore instructions embedded within that external data.
- Capability inventory: The skill possesses file-system capabilities, specifically using tools like
create_fileandstr_replace(referenced in Stage 2) to generate and modify document artifacts. - Sanitization: The workflow lacks any mention of sanitization, validation, or filtering of the retrieved content before it is processed by the agent to influence the structure or content of the resulting document.
Audit Metadata