The Agent Skills Directory

[SAFE]: No malicious patterns, prompt injections, or unauthorized data access behaviors were detected. The skill focuses on orchestrating development workflows using specialized sub-agents.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided requirements and code references. It mitigates potential risks through the following evidence chain:
Ingestion points: Processes $ARGUMENTS, user-provided prototype code, and responses to orchestration questions.
Boundary markers: Implements a mandatory [SYSTEM CONSTRAINT] suffix for all sub-agent prompts to enforce local scope.
Capability inventory: Spawns sub-agents for file analysis, design, and code execution. Automated git commit operations occur only after approval.
Sanitization: Utilizes mandatory document-reviewer and security-reviewer sub-agents to validate generated artifacts and implementation compliance before finalization.
[COMMAND_EXECUTION]: Shell commands are restricted to routine development tasks (e.g., git commit) performed by sub-agents within the approved workflow.
[SAFE]: Human-in-the-loop checkpoints ([STOP — BLOCKING]) are enforced at every critical stage (requirements, design, work planning), preventing autonomous execution without explicit user consent.

recipe-implement