The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze user-provided legal judgment documents, which constitutes an ingestion point for untrusted data. This creates a surface for indirect prompt injection where a malicious document could attempt to influence the agent's logic. Ingestion points: Judgments provided as input as specified in SKILL.md and extraction guidelines in references/judgment-first-instance-extraction.md. Boundary markers: SKILL.md and the provided templates (e.g., templates/template-internal.md) enforce strict structural boundaries, instructing the agent to differentiate between court认定, party claims, and its own analytical observations. Capability inventory: No subprocess calls, network operations, or dynamic code execution capabilities are present; the skill only generates text and Mermaid diagrams. Sanitization: Instructions require the agent to stick strictly to the text of the provided document and explicitly label missing information rather than making assumptions or filling in gaps.- [SAFE]: No malicious patterns such as obfuscation, command execution, or data exfiltration were detected. The skill relies entirely on standard markdown instructions and Mermaid syntax for visualization. All external links point to the author's official GitHub repository or established licensing organizations.

judgment-analysis