The Agent Skills Directory

[PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection because it instructs the agent to ingest and evaluate content from external legal databases and third-party lawyer articles.
Ingestion points: External data is fetched via tools referencing [YD] 法律精灵, [GOV] 裁判文书网, and [WKL] 律师分析 as specified in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters provided to prevent the agent from executing instructions potentially embedded in the fetched legal documents or articles.
Capability inventory: The skill is purely informational and lacks any identified capabilities for shell execution, file system modification, or network-based exfiltration.
Sanitization: No sanitization or validation protocols are established for the external data before it is processed by the agent.
[SAFE]: The skill contains no obfuscated code, hardcoded credentials, or unauthorized persistence mechanisms. Accessing context from ../CLAUDE.md is a standard pattern for shared configurations in this environment and does not pose a security risk in this context.

region-adjudicator