The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates external workspace data and user-provided responses into its generation process without explicit sanitization.
Ingestion points: The skill reads vision and business-model files from the local workspace and processes unstructured data from user interviews (SKILL.md).
Boundary markers: The instructions do not specify the use of clear delimiters or guardrails to prevent instructions embedded within the vision or business-model documents from influencing the agent's behavior.
Capability inventory: The skill is capable of writing multiple files (OPERATIONS.md, operations.json, CLAUDE.md), updating database tables (decisions, meetings), and invoking MCP tools for querying and recording data.
Sanitization: No explicit input validation or escaping mechanisms are defined for the content extracted from external documents.
[DYNAMIC_EXECUTION]: The skill generates structured configuration files, database schema definitions, and a document-based website at runtime.
Evidence: The instructions direct the agent to "ship the result as a Business OS module," which involves creating a machine contract (operations.json), operational tables, MCP tool configurations, and a "single-page-scroll document site" using brand-specific tokens (SKILL.md).

operations