internal-comms
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and summarize information from potentially untrusted external sources like Slack, Google Drive, and email.
- Ingestion points: Multiple guideline files (examples/3p-updates.md, examples/company-newsletter.md, examples/faq-answers.md) direct the agent to search through Slack channels, internal emails, calendar events, and shared documents to gather content for reports and newsletters.
- Boundary markers: The skill lacks instructions for the agent to use delimiters or ignore embedded instructions when reading from these external sources, increasing the risk of the agent following malicious prompts hidden in the retrieved data.
- Capability inventory: The skill leverages the agent's ability to read communication and document metadata to construct internal updates. A malicious message or document could influence the tone, content, or further actions of the agent.
- Sanitization: No explicit sanitization, filtering, or validation steps are defined for the content extracted from these third-party platforms before it is drafted into communication materials.
Audit Metadata