The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (agendas, board presentations, management reports) to generate minutes, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: In Step 3, the skill requests the user to provide agenda items, pre-read materials, board presentations, and management reports for extraction.
Boundary markers: Absent. There are no instructions to the agent to treat the content of these materials as data only or to ignore embedded instructions.
Capability inventory: The skill has the capability to read from and write to the local filesystem at paths defined in ~/.claude/plugins/config/claude-for-legal-zh/.
Sanitization: Absent. The skill summarizes and interpolates the content of these materials directly into the minute drafts without validation or escaping.
[DATA_EXFILTRATION]: The skill accesses local configuration files that contain sensitive corporate and legal information.
Evidence: The skill reads ~/.claude/plugins/config/claude-for-legal-zh/corporate-legal/CLAUDE.md and matter.md files to retrieve board member lists, committee details, and matter-specific context.

board-minutes