The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted external data such as customer problem statements, RFPs, and discovery notes (defined in SKILL.md Phase 1.1) to drive the subsequent autonomous build and costing phases.
Ingestion points: Phase 1.1 Intake & Parsing processes raw customer briefs and RFPs.
Boundary markers: The skill uses an 'Assumption Log' (Phase 1.2) and a 'Pre-Build Gate' (Phase 4.1) to validate scope before execution, though it lacks specific delimiters for natural language instructions.
Capability inventory: The skill generates and executes full-stack application code, running tests and deployment commands via npm, pytest, and docker-compose (SKILL.md Phase 4.6).
Sanitization: The skill advocates for strict schema validation (Zod, Pydantic) within the generated applications (Phase 4.3), but does not explicitly describe sanitization for the initial natural language ingestion phase.
[DYNAMIC_EXECUTION]: The skill operates as an autonomous build engine (Phase 4) that generates repository structures, backend logic (Node.js/Python), and frontend components (React/Next.js) from scratch. It then instructs the agent to execute these generated scripts through test runners and container orchestration tools.
[COMMAND_EXECUTION]: The workflow involves the execution of various system commands for environment setup, dependency management, and quality assurance, including docker-compose, npm test, pytest, and terraform (SKILL.md Phase 4.6 and infrastructure standards).

proposal-automation