internal-comms-anthropic

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to process content from third-party-controlled sources like Slack, Email, and Google Drive.
  • Ingestion points: The files examples/3p-updates.md, examples/company-newsletter.md, and examples/faq-answers.md instruct the agent to gather information from Slack, Google Drive, Email, and Calendar.
  • Boundary markers: There are no instructions to use delimiters (e.g., XML tags or hashes) or specific directives to ignore potential commands embedded in the source data.
  • Capability inventory: The agent acts as a summarizer and drafter of communications based on the gathered data, which could be influenced by adversarial instructions in the input.
  • Sanitization: No methods for validating or sanitizing the content retrieved from external tools are mentioned.
  • [NO_CODE]: This skill consists entirely of Markdown prompt engineering and configuration; it does not contain any executable scripts (Python, JavaScript, etc.), which significantly reduces the risk of direct malicious actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 02:58 PM