Agent Skills › yielded-dev/auth › feedback-loops

feedback-loops

GitHub

优化开发反馈循环,设计可复用的测试夹具、基准测试和验证命令。解决检查缓慢、重复或不可靠问题,提升获取可信答案的速度,通过改进工具链和隔离资源实现高效迭代。

.agents/skills/feedback-loops/SKILL.md yielded-dev/auth

Trigger Scenarios

测试或检查运行缓慢 需要建立可复用的验证流程 CI/CD 中的质量门禁配置 消除重复或不可靠的检查

Install

npx skills add yielded-dev/auth --skill feedback-loops -g -y
More Options

Non-standard path

npx skills add https://github.com/yielded-dev/auth/tree/main/.agents/skills/feedback-loops -g -y

Use without installing

npx skills use yielded-dev/auth@feedback-loops

指定 Agent (Claude Code)

npx skills add yielded-dev/auth --skill feedback-loops -a claude-code -g -y

安装 repo 全部 skill

npx skills add yielded-dev/auth --all -g -y

预览 repo 内 skill

npx skills add yielded-dev/auth --list

SKILL.md

Frontmatter
{
    "name": "feedback-loops",
    "description": "Design or improve reusable development feedback loops, benchmark harnesses, and verification commands when checks are slow, duplicated, unreliable, or missing."
}

Fast, reusable feedback loops

Optimize time to a trustworthy answer. Name the behavior, invariant, or metric that decides the next step and run the smallest check that answers it. Reuse valid evidence. Broaden or repeat checks for changed inputs, a failure, or an unresolved risk; do not verify every action independently.

Use the existing app or library harness. When a recurring question lacks one, improve its owner's tooling and share the real operation across agent commands, tests, and CI. Small exploratory probes are useful; promote recurring measurements into the harness. Keep raw run evidence in artifacts.

  • Building a harness, CI gate, or durable check for a recurring mistake: harness design.
  • Selecting cases, caching fixtures, or fixing slow setup, queues, and failures: iteration cost.

Use isolated Cloudflare resources when they shorten feedback or provide the actual runtime. Their creation and teardown through Alchemy are pre-authorized; temporary stacks owns isolation, reuse, and cleanup.

Never build or wait for a laptop-wide CPU lock or an idle machine. Local deterministic checks need no exclusivity. Move contention-sensitive timing to isolated cloud resources and run independent jobs concurrently within actual provider/project limits. Account for cloud noise too.

For traces and attribution, select observability. For comparable measurements and performance acceptance, use measure-performance; reuse already loaded guidance.

Version History

  • 72bfeeb Current 2026-10-18 15:34

Same Skill Collection

.agents/skills/alchemy/SKILL.md
.agents/skills/cloudflare/SKILL.md
.agents/skills/dev-kit/SKILL.md
.agents/skills/effect/SKILL.md
.agents/skills/library-first/SKILL.md
.agents/skills/measure-performance/SKILL.md
.agents/skills/o11y/SKILL.md
.agents/skills/project-setup/SKILL.md
.agents/skills/pull-requests/SKILL.md
.agents/skills/show-me/SKILL.md
.agents/skills/simplify/SKILL.md
.agents/skills/testing/SKILL.md

Metadata

Files
0
Version
72bfeeb
Hash
3d665b5a
Indexed
2026-10-18 15:34

홈 - 위키
Copyright © 2011-2026 iteam. Current version is 2.155.2. UTC+08:00, 2026-10-19 11:11
浙ICP备14020137号-1