meeting-briefing-anthropic

Warn

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: MEDIUMPROMPT_INJECTION
Full Analysis
  • [METADATA_POISONING]: The skill's metadata identifies the author as "Anthropic", which directly contradicts the system-provided author context of "lawve-ai". This discrepancy is deceptive and may lead users to incorrectly trust the skill's origin and safety profile.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, creating a significant attack surface for indirect prompt injection.
  • Ingestion points: The skill is instructed to gather context from Emails, Chat messages (Slack/Teams), and external Documents.
  • Boundary markers: There are no instructions defining delimiters or clear boundaries to isolate ingested data from the system prompt, nor are there warnings to ignore embedded instructions.
  • Capability inventory: The skill aggregates and synthesizes sensitive business information and possesses the capability to distribute results and action items via email and other channels.
  • Sanitization: The methodology does not include any validation, filtering, or sanitization steps for the data retrieved from external sources before it is processed and redistributed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 16, 2026, 04:13 AM
Security Audit — agent-trust-hub — meeting-briefing-anthropic