The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill serves as a pure dispatch primitive, mapping user keywords and intent to internal sub-skills.
[PROMPT_INJECTION]: The skill processes natural language input to determine routing logic (Indirect Prompt Injection surface).
Ingestion points: User requests are parsed for keywords, file types, and verbs in SKILL.md (Step 1).
Boundary markers: None explicitly defined; the skill relies on LLM classification of the user's request.
Capability inventory: The skill is limited to producing dispatch decisions and does not execute code, access sensitive files, or perform network requests.
Sanitization: Not present at the routing layer, as the skill's output is restricted to predefined target names and descriptive briefs.
[COMMAND_EXECUTION]: The skill references and dispatches to various sub-skills (e.g., legal-contract-review, legal-playbook-author). These are identified as vendor-owned resources within the legal-toolkit ecosystem and represent standard tool-chaining functionality.

using-legal-toolkit