The Agent Skills Directory

Indirect Prompt Injection: The skill processes information from untrusted document sources, creating a potential surface for indirect instruction injection.
Ingestion points: Litigation documents including PDFs, emails, .docx files, and eDiscovery exports from platforms like Everlaw or Relativity.
Boundary markers: The skill incorporates a structured system of reviewer notes and flags, such as [VERIFY], [CITE NEEDED], and [SME VERIFY], to ensure extracted content is validated by a human user.
Capability inventory: The skill possesses the ability to read and write files within specific matter directories and utilize MCP connectors for querying external document storage systems.
Sanitization: The skill does not explicitly mention sanitization of the content extracted from documents before it is formatted into the markdown timeline.
Matter Workspace Security: The skill enforces a mandatory conflicts check by verifying entries in a local _log.yaml file before building any work product. This prevents the unauthorized creation of chronologies for matters that have not undergone formal intake procedures.

chronology