internal-comms

Pass

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection.
  • Ingestion points: Instructions in 'examples/3p-updates.md', 'examples/company-newsletter.md', and 'examples/faq-answers.md' direct the agent to read content from Slack, Google Drive, Email, and Calendar.
  • Boundary markers: The skill fails to use delimiters or 'ignore embedded instructions' warnings when processing this external data.
  • Capability inventory: The skill utilizes the agent's broad read access to integrated corporate communication and storage tools.
  • Sanitization: There are no instructions to validate or sanitize the ingested data, allowing maliciously crafted Slack messages or documents to potentially hijack the agent's behavior.
  • [DATA_EXFILTRATION]: The skill facilitates the aggregation of sensitive corporate data.
  • Evidence: It directs the agent to search for 'emails from executives,' 'critical team member docs,' and 'company-wide vision docs.' If the agent is successfully manipulated via injection, this sensitive information could be leaked into the final communication outputs.
  • [NO_CODE]: The skill consists entirely of markdown instructions and lacks any scripts or executable code.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 26, 2026, 04:44 AM