The Agent Skills Directory

[SAFE]: The skill acts as a logical router for legal tasks and does not perform any high-risk operations such as network requests, file writes, or command execution. It uses a restricted platform configuration that limits tool access to Read and ensures data residency remains local.
[SAFE]: Data protection is a central design element; the instructions mandate the use of a pseudonymization skill (let-it-be) before processing any sensitive data or personally identifiable information (PII). This mitigates risks associated with processing legal content.
[SAFE]: The skill provides structured output (JSON) and clear decision tables, which reduces the risk of logic errors or unintentional behavior during the triage process.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as it processes untrusted legal queries to determine routing logic.
Ingestion points: Legal queries and deliverables are ingested via the agent's context as described in SKILL.md.
Boundary markers: The skill uses a strict JSON schema for output to ensure predictable downstream behavior, though it lacks explicit delimiters for user input in the prompt logic.
Capability inventory: Access is restricted to the Read tool; no capabilities for file writing, network operations, or dynamic code execution are present.
Sanitization: Explicitly instructs the agent to use the let-it-be skill for pseudonymization of PII before further processing.

legal-request-router-pl