The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project-specific documentation to guide its behavior, creating an indirect prompt injection surface.\n
Ingestion points: The skill reads from IMPLEMENTATION_PLAN.md, PRD.md, progress.txt, TECH_STACK.md, DESIGN_SYSTEM.md, FRONTEND_GUIDELINES.md, and BACKEND_STRUCTURE.md as outlined in SKILL.md and references/session-startup.md.\n
Boundary markers: The skill requires an implementation plan to be written to tasks/todo.md and verified by the user before code is written, as described in references/planning.md. It also treats the documentation suite as a locked authority to prevent unauthorized deviations.\n
Capability inventory: The skill possesses file system access for creating and modifying code, and uses specialized tools like vibe-coding-cli-runner for executing development commands as shown in SKILL.md.\n
Sanitization: Mandatory code reviews via vibe-coding-code-review and specific testing requirements in references/engineering-standards.md serve as safety checkpoints to validate the integrity of the output.\n- [COMMAND_EXECUTION]: The skill utilizes a specialized tool to perform necessary development tasks.\n
Evidence: The vibe-coding-cli-runner tool is invoked for CLI-driven tasks such as migrations, code generation, and scaffolding during the step execution phase in SKILL.md. This is a legitimate development capability managed through the agent's tool-execution interface and scoped to the build process.

vibe-coding-build