dingtalk-minutes
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script 'scripts/minutes_recent_summary.py' utilizes the subprocess module to execute the 'dws' binary, which is a required tool for fetching and managing recording data. The command construction uses list-based arguments without a shell, which is a standard and secure implementation for local CLI wrappers.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes potentially untrusted external data in the form of meeting transcripts and AI-generated summaries.
- Ingestion points: Meeting content is retrieved from the DingTalk platform via 'dws' commands in 'scripts/minutes_recent_summary.py' and referenced in 'SKILL.md'.
- Boundary markers: The 'DWS_RUNTIME_CONTRACT' in 'SKILL.md' mandates the use of structured JSON for reading and explicitly forbids guessing identifiers, reducing the risk of unintended instruction execution.
- Capability inventory: The skill possesses capabilities for command execution (via 'dws') and local file system writes (for exporting pack files).
- Sanitization: The skill documentation for '+export-pack' in 'references/intent-guide.md' describes a recursive sanitization process that replaces recognized signature URLs and credentials with redaction markers before the data is saved locally.
Audit Metadata