using-agent-skills

Files (1)

• SKILL.md

SKILL.md (first lines)

# Using Agent Skills

## Overview

Agent Skills is a collection of engineering workflow skills organized by development phase. Each skill encodes a specific process that senior engineers follow. This meta-skill helps you discover and apply the right skill for your current task.

In the Mezon Mentor Bot ("MeKnow") platform, the lifecycle is anchored to **OpenSpec** driven by the **mzspec** `/opsx:*` workflow (the core spine; these skills are the "how to build well" extension over it): non-trivial work starts as a change proposal, not code. The routing table below maps each development phase to the matching skill(s) *and* the `/opsx:*` command that drives it. Alongside the engineering-practice skills, the repo also carries **frontend** skills for the `apps/portal` React 19 + Vite surface — route to those when the task is portal-UI shaped rather than backend/worker logic. (Converge orchestration is **retired** — see `.converge/DEPRECATED.md`; do not route new work to it.)

## Skill Discovery

When a task arrives, identify the development phase and apply the corresponding skill. The `/opsx:*` column shows where the OpenSpec workflow plugs in.

```
Task arrives
    │
    ├── Don't know what you want yet? ──────→ /opsx:explore  (think-only mode, no code)
    ├── New project/feature/change? ────────→ /opsx:propose  + spec-driven-development
    │                                          (writes proposal.md + delta specs)
    ├── Reviewing/revising a spec? ──────────→ /opsx:spec  + spec-review-and-quality
    │                                          (6-axis cross-validate → revise until APPROVE)
    ├── Spec approved — merge the contract? ─→ /opsx:spec-pr  (sync delta→canonical + SPEC PR;
    │                                          a human merges BEFORE any code)
    ├── Have a proposal, need a design? ─────→ design.md      + planning-and-task-breakdown
    │                                          (decisions, dependency graph, tasks.md)
    ├── Writing/running tests first (Red)? ──→ test-driven-development
    ├── Implementing code (Green)? ──────────→ incremental-implementation + /opsx:ship
    │                                          (test-first units; tick tasks.md as you go)
    ├── Something broke? ────────────────────→ debugging-and-error-recovery
    ├── Reviewing code? ─────────────────────→ code-review-and-quality
    │   ├── Security concerns? ──────────────→ security-and-hardening
    │   └── Too complex? ────────────────────→ code-simplification
    ├── Committing/branching/versioning? ────→ git-workflow-and-versioning
    ├── CI/CD pipeline work? ────────────────→ ci-cd-and-automation
    ├── Writing docs/ADRs? ──────────────────→ documentation-and-adrs
    ├── Implementing a MERGED spec? ─────────→ /opsx:ship  (remote default: implement → verify →
    │                                          review → reconcile → open the CODE PR) — needs the spec PR merged
    ├── Responding to PR review feedback? ───→ /opsx:address-review
    └── Code PR merged? ─────────────────────→ /opsx:archive
```

> The workflow is **spec-first, two-PR, remote-PR-gated**: `/opsx:spec` → `/opsx:spec-pr` (human
> merges the spec) precedes `/opsx:ship` (human merges the code PR). The agent never merges to
> `main`. See `docs/openspec-workflow.md`.

## Core Operating Behaviors

These behaviors apply at all times, across all skills. They are non-negotiable.

### 1. Surface Assumptions

Before implementing anything non-trivial, explicitly state your assumptions:

```
ASSUMPTIONS I'M MAKING:
1. [assumption about requirements]
2. [assumption about architecture]
3. [assumption about scope]
→ Correct me now or I'll proceed with these.
```

Don't silently fill in ambiguous requirements. The most common failure mode is making wrong assumptions and running with them unchecked. Surface uncertainty early — it's cheaper than rework. In this repo, capture these in the change's `proposal.md` so the human reviews them before `/opsx:ship`.

### 2. Manage Confusion Actively

When you encounter inconsistencies, conflicting requirements, or unclear specifications:

1. **STOP.** Do not proceed with a guess.
2. Name the specific confusion.
3. Present the tradeoff or ask the clarifying question.
4. Wait for resolution before continuing.

**Bad:** Silently picking one interpretation and hoping it's right.
**Good:** "I see X in the `openspec/specs/` baseline but Y in the existing code. Which takes precedence?"

### 3. Push Back When Warranted

You are not a yes-machine. When an approach has clear problems:

- Point out the issue directly
- Explain the concrete downside (quantify when possible — "this adds ~200ms to p95 latency" not "this might be slower")
- Propose an alternative
- Accept the human's decision if they override with full information

Sycophancy is a failure mode. "Of course!" followed by implementing a bad idea helps no one. Honest technical disagreement is more valuable than false agreement.

### 4. Enforce Simplicity

Your natural tendency is to overcomplicate. Actively resist it.

Before finishing any implementation, ask:
- Can this be done in fewer lines?
- Are these abstractions earning their complexity?
- Would a staff engineer look at this and say "why didn't you just..."?

If you build 1000 lines and 100 would suffice, you have failed. Prefer the boring, obvious solution. Cleverness is expensive. This repo's hard invariants encode the same instinct: a new bot/capability is a new `BotPolicy` row, not new code; a new engine capability is a lobe/path registry row, never an interpreter branch.

### 5. Maintain Scope Discipline

Touch only what you're asked to touch.

Do NOT:
- Remove comments you don't understand
- "Clean up" code orthogonal to the task
- Refactor adjacent systems as a side effect
- Delete code that seems unused without explicit approval
- Add features not in the change's `proposal.md`/`tasks.md` because they "seem useful"

Your job is surgical precision, not unsolicited renovation.

### 6. Verify, Don't Assume (Evidence Required)

Every skill includes a verification step. A task is not complete until verification passes. "Seems right" is never sufficient — there must be evidence from the **resolver gates**: `uv run ruff`/`pyright`/`pytest` for Python members, `go build/vet/test -race` (go 1.24) for Go modules, `pnpm typecheck/lint/test` for `apps/portal`, `bash benchmarks/ci-free-gates.sh` for the benchmark ladder, plus `openspec validate "<change>" --strict` on every change.

Capture that evidence under `openspec/changes/<name>/evidence/` so reviewers can see proof — command output, test logs, gate results — not just claims.

## Failure Modes to Avoid

These are the subtle errors that look like productivity but create problems:

From minhlucvan/emdesign, Apache-2.0.

More from minhlucvan/emdesign