doc-coauthoring
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its document ingestion workflow.
- Ingestion points: The agent is instructed to read content from external shared documents via links and messaging platforms like Slack and Teams through connectors.
- Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' directives for the ingested content.
- Capability inventory: The skill utilizes tools such as create_file and str_replace and has the capability to spawn sub-agents for reader testing.
- Sanitization: There is no requirement in the skill for input validation or sanitization of data retrieved from these external sources.
- [COMMAND_EXECUTION]: The skill uses filesystem tools like create_file and str_replace to manage document drafts. While these are necessary for the skill's functionality, they operate on data derived from potentially untrusted external sources, creating a risk of unintended file modifications if an indirect prompt injection occurs.
Audit Metadata