Add Agent Task Gates — AI employee task-discipline agent

[kongminOS]

Adds agents/agent-task-gates.agent.md — a task-discipline gate keeper for
AI-employee and multi-agent teams.

It enforces a four-gate pipeline before work is assigned and before work is
accepted:

• G1 start-up review (grill-me): context self-check before any task begins
• G2 requirements-to-spec: six-section spec with explicit out-of-scope
• G3 spec-to-tickets: vertical slices with edge-blocking declarations
• G4 implement-and-verify: TDD, per-file type checks, full suite, code review

Why this slot: the existing agent-governance-reviewer agent covers security
and trust governance (policies, intent classification, audit trails). This
agent covers the other half — task-execution discipline and acceptance: what
happens before an agent starts and before its work is accepted. Both are
needed to run AI teams in production.

The discipline adapts mattpocock's production-grade skill system (grill-me,
to-spec, to-tickets, implement) to AI-employee governance, with machine
enforcement and KPI accountability. Source credit is included in the agent
body. Frontmatter follows the CONTRIBUTING agent format (description / model /
tools / name) and the filename follows the lowercase-hyphen .agent.md
convention. I could not run npm run build locally to regenerate the README
tables — happy to adjust if the format needs anything else.