The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script, scripts/graph-audit, to perform integrity checks on durable files after completing a roadmap step.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from multiple user-managed files, creating a potential surface for indirect prompt injection.
Ingestion points: The skill reads from baton.md, wiki/business/vision.md, initiative files within wiki/initiatives/, and user-provided inputs in raw/inbox/.
Boundary markers: The skill relies on citing sources and factual reporting but does not implement explicit structural boundaries or delimiters for all data ingested from the vault files.
Capability inventory: The skill possesses capabilities for local file system read/write operations and the execution of the scripts/graph-audit command.
Sanitization: The workflow incorporates a 'shared durable claim admission gate' at write time and utilizes the graph-audit script to enforce data relationships and verify integrity.

sprint