lark-minutes

Warn

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [METADATA_POISONING]: The skill frontmatter claims the author is 'larksuite' and provides a source URL for the official 'larksuite/cli' repository, despite the skill being provided by 'seaworld008'. This impersonation is a deceptive practice. Additionally, the metadata contains a future creation date ('2026-05-19'), which is inconsistent and misleading.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from meeting transcripts and summaries.
  • Ingestion points: Untrusted meeting content is read via the +detail command (referenced in SKILL.md).
  • Boundary markers: The skill does not define or use security delimiters to separate untrusted meeting content from instructions.
  • Capability inventory: The agent is granted capabilities to modify meeting data (+summary, +todo, +update) and perform local file writes (+download) based on processed inputs.
  • Sanitization: There is no requirement or logic for sanitizing or validating external content before it is processed by the agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 18, 2026, 12:23 PM
Security Audit — agent-trust-hub — lark-minutes