The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation specifies the execution of local shell scripts and system utilities for document processing. This includes validate_brd_wrapper.sh, validate_all.py, and standard commands like sha256sum, grep, and ls. These commands are used to enforce document standards and verify file integrity within the project's local directory structure.\n- [PROMPT_INJECTION]: As the skill processes various external inputs such as reference documents, implementation plans, and direct user prompts, it contains an attack surface for indirect prompt injection. This risk is managed through a multi-phase validation process (Phases 4-5) that leverages a unified quality gate (doc-brd-audit) and an automated fixer (doc-brd-fixer) to ensure adherence to strict markdown templates and YAML schemas. Mandatory Evidence Chain: 1. Ingestion points: docs/00_REF/, REF/, work_plans/, and user prompts via --prompt. 2. Boundary markers: Explicit schema validation (BRD_MVP_SCHEMA.yaml) and template compliance checks (BRD-MVP-TEMPLATE). 3. Capability inventory: Local bash and python script execution, file reading, and updating of index documents. 4. Sanitization: Iterative review/fix cycles with scoring gates and template alignment checks.

doc-brd-autopilot