The Agent Skills Directory

[SAFE]: The skill contains no executable scripts, binaries, or shell commands. It is entirely composed of markdown instructions for text processing.
[SAFE]: No external URLs, remote downloads, or network operations are present in the instructions.
[SAFE]: There are no hardcoded credentials, API keys, or attempts to access sensitive system files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user data in the form of legal briefs.
Ingestion points: Brief text, case metadata, and local rules are ingested as the primary data source (SKILL.md).
Boundary markers: The skill explicitly instructs the agent to use [__] or [VERIFY] placeholders for any data that cannot be verified verbatim from the source.
Capability inventory: No capabilities for code execution (eval/exec), file system writing, or network communication were identified.
Sanitization: While no technical sanitization is mentioned, the skill mandates that a licensed attorney review all outputs, which serves as a human-in-the-loop safety measure.

appellate-formatting