跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
其他技能

using-a-team

The meta-skill. Loaded at every session start. Defines which A Team skills and agents MUST be used for which situations. If a skill or agent applies, you do not have a choice — you must use it.

RBraga01 · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add RBraga01/a-team --skill using-a-team

技能说明


name: using-a-team description: The meta-skill. Loaded at every session start. Defines which A Team skills and agents MUST be used for which situations. If a skill or agent applies, you do not have a choice — you must use it.

Using A Team

This skill is injected at every session start. It defines mandatory skill and agent usage.

Before ANY Code Is Written

SituationRequired Skill/Agent
New feature, component, or capabilitybrainstorming skill first
Have a spec, need an implementation planwriting-plans skill
Have a plan, executing itsubagent-driven-development or executing-plans skill
Starting development workusing-git-worktrees skill (create isolation first)

During Development

SituationRequired Skill/Agent
Writing any new function or featuretest-driven-development skill (RED before GREEN)
Any bug, test failure, unexpected behaviordebugger agent (systematic debugging — Phase 1 first)
Multiple independent problemsdispatching-parallel-agents skill
Build or type errorsbuild-error-resolver agent (minimal diffs only)

Before Claiming Completion

ClaimRequired Skill
"The tests pass"verification-before-completion (run tests, read output)
"The build succeeds"verification-before-completion (run build, read output)
"The bug is fixed"verification-before-completion (reproduce fix, read output)
"The feature works"verification-before-completion (run the feature, read output)
"Done" / "Complete"verification-before-completion (ALWAYS, no exceptions)

After verification passes, Step 5 of verification-before-completion is mandatory: delete all log files, stack traces, and debug dumps generated during this task before reporting done.

After Code Changes

SituationRequired Agent
After writing or modifying codecode-reviewer agent
Auth, API, input handling, DB changessecurity-reviewer agent
Before any PR mergequality-gate command (runs both)
Wrapping up a branchfinishing-a-development-branch skill

Architectural Decisions

SituationRequired Agent
System design, new feature architecturearchitect agent (produces ADR)
Complex feature planningplanner agent
Tech stack or pattern decisionarchitect agent
Inherited/unfamiliar codebase, scaling milestone, or periodic health checkarchitecture-audit skill

Language & Domain Reviews

SituationRequired Agent
Go files changedgo-reviewer
Python files changedpython-reviewer
Rust files changedrust-reviewer
Kotlin / Android files changedkotlin-reviewer
Swift / iOS files changedswift-reviewer
Dart / Flutter files changedflutter-reviewer
SQL / migrations / schema changeddatabase-reviewer
Terraform / Docker / K8s / CI changedinfra-reviewer
Any LLM API calls added or changedai-reviewer
Any privacy / payment / child data codecompliance-reviewer

Before Any API Endpoint

SituationRequired Skill
Writing a new REST / gRPC / GraphQL / event endpointapi-contract-first (write the contract first)
Any ALTER TABLE, DROP, or backfill in productiondata-migration (rollback plan first)

Performance & Production

SituationRequired
Performance regression reportedperformance-profiler agent (measure first)
Performance-critical feature pre-releaseperformance-audit skill
Production is degraded or downincident-response skill (immediately)

During Code Changes — Surgical Changes Rule

Every change must be scoped to what was requested. Before committing:

ViolationWhat it looks like
Drive-by refactoringRenamed variables, added type hints, changed quote style — none of it was in the task
Style normalizationRewrote adjacent code to match your preferences instead of the existing style
Orphan overreachRemoved dead code that existed before this task (only YOUR orphans are your responsibility)

Self-check: "Would a reviewer see any line in this diff that cannot be explained by the task description?" If yes → undo it.

Rationalization Red Flags

These thoughts mean you are about to skip a mandatory step. Stop.

  • "This change is too small to need a review"
  • "I'll write the test after I confirm it works"
  • "The plan is in my head, I don't need to write it"
  • "The tests were passing before so they're probably still passing"
  • "I know this language well enough, no need for the reviewer"
  • "The skill overhead slows me down" (it doesn't — it prevents the rework that slows you down)

Session Start Checklist

On every session start, confirm:

  • Is INIT.md present? If not → fill out INIT_TEMPLATE.md first
  • Is .agent-sync/TEAM.md present? If not → run /orchestrate init first
  • What is the current task? → identify which skills/agents apply before starting

Skill Registration (For New Skills)

When a new skill is added to A Team, add it to the trigger table above so it gets enforced from day one.

The A Team is only as good as the discipline with which it is applied. Skills consulted are skills that work. Skills skipped are skills that don't exist.