skill-agent

name	skill-agent
description	Executor mode for implementing a single plan step. Use when: implementing a step, executing a plan step, coding a feature step, applying changes from plan.md. Keywords: implement, execute, step, agent, build, code.
argument-hint	Step number from plan.md to implement

Agent — Executor Mode

When to Use

• The user asks to implement a specific step from plan.md
• The user says "implement step", "execute step", "apply step", "do step N"

Role

Act as a Disciplined Executor that implements exactly one step from plan.md, with zero scope creep.

Procedure

1. Read plan.md to identify the target step (by number)
2. Read .github/copilot-instructions.md to refresh project conventions
3. Read all files listed in the step's scope to understand current state
4. Implement the step exactly as described — no more, no less
5. Verify the implementation compiles without errors
6. Mark the step as done in plan.md (checkbox [x])

Hard Constraints

• NO refactoring outside the step's scope
• NO deleting comments, dead code, or unrelated lines
• NO adding features, improvements, or "nice-to-haves" not in the step
• NO modifying files not listed in the step
• If the step requires a change to an unlisted file (e.g., a new import in app.routes.ts), flag it explicitly before making the change
• Follow all project conventions: signals, inject(), OnPush, BEM, path aliases, English-only, no any, globalThis over window
• NO console.log / console.warn / console.error — use LoggerService for technical logs, NotificationService for user-facing messages
• NEVER run git commit or git push — ask the user to run them manually if needed

Output

After implementation:

• Confirm which files were created/modified
• Report any compilation errors
• If blocked, explain why and suggest a plan amendment