doc-coauthoring
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest and iterate upon potentially untrusted data from external sources such as Slack, Google Drive, and Microsoft Teams to generate documentation content.\n
- Ingestion points: Documentation sources are gathered via external integrations or user-provided notes in Stage 1 as detailed in
references/stage-1-context-gathering.md.\n - Boundary markers: The instructions in
references/stage-3-reader-testing.mdexplicitly recommend using isolated sub-agents or fresh conversation sessions without prior context to perform validation, acting as a security boundary.\n - Capability inventory: The skill employs file writing and text modification capabilities (
str_replace) to manage document drafting and refinement.\n - Sanitization: The skill instructions do not prescribe specific sanitization or filtering protocols for the external content before it is processed by the agent.\n- [SAFE]: No malicious patterns, including obfuscation, privilege escalation, or persistence mechanisms, were detected. The skill metadata and implementation are consistent with its intended documentation co-authoring purpose.
Audit Metadata