lab-meeting-report

Pass

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external and potentially untrusted sources, including PDF papers, Notion pages, and Lark chat messages.
  • Ingestion points: Detailed in SKILL.md (Step 1), references/notion-integration.md, and references/lark-integration.md.
  • Boundary markers: The references/notion-integration.md file contains explicit security warnings and instructions to disregard any embedded commands found within the data.
  • Capability inventory: The skill can execute shell commands via tools like lark-cli, marp, and quarto, and has local file write access.
  • Sanitization: Implements a multi-step evidence verification process, including an internal ledger and quality gate, to ensure report accuracy and grounding.
  • [COMMAND_EXECUTION]: The skill uses command-line interfaces to handle cloud synchronization and document conversion.
  • Evidence: Mentions and utilizes lark-cli for Feishu/Lark integration, and marp or quarto for generating presentation files.
  • [EXTERNAL_DOWNLOADS]: The skill refers to several well-known academic and presentation software tools.
  • Evidence: References to Marp CLI, Quarto, Slidev, and PptxGenJS for creating derived presentation artifacts.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 24, 2026, 06:50 AM