The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file contains a dynamic context placeholder using the !command syntax to execute a shell command (skill-toc) at load time. This is a benign utility for generating a table of contents and does not involve unauthorized file access or data exfiltration.- [COMMAND_EXECUTION]: The skill instructions include executing the bundled Python script, scripts/westlaw-docx-text.py, to process downloaded legal documents. An analysis of the script confirms it uses only Python standard libraries and performs safe XML parsing via regex on local files.- [INDIRECT_PROMPT_INJECTION]: The skill processes external DOCX files downloaded from Westlaw, creating an ingestion surface for untrusted data. Evidence chain: 1. Ingestion point: Westlaw DOCX files retrieved during automation. 2. Boundary markers: None explicitly defined. 3. Capability inventory: Local script execution and browser automation via CDP. 4. Sanitization: The text extraction script performs structural parsing without executing or evaluating the document content, mitigating the risk of injection affecting the agent.

westlaw