← 返回 Skills 市场
modeyapu

Continuous Dev Loop

作者 ModeYapu · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ✓ 安全检测通过
47
总下载
0
收藏
1
当前安装
2
版本数
在 OpenClaw 中安装
/install continuous-dev-loop
功能描述
单-slice 持续开发循环 V2:更强验证、抗卡顿恢复、明确 Next 接力。
使用说明 (SKILL.md)

Continuous Dev Loop

Use when the user wants a code project driven in repeated, bounded engineering rounds instead of ad-hoc task hopping.

Use This For

  • Ongoing development where one verified slice should land before the next begins.
  • Repositories where continuity should come from code, docs, and prior validation evidence rather than chat memory alone.
  • Sessions where each round should end with a clean handoff for the next round.

V2 Core Rules

  • Treat one round as exactly: development -> verification -> optimization -> planning.
  • Each round must pursue exactly one slice. Do not branch into a second feature.
  • Continue from the last verified repo state and explicit repo evidence first, not conversational recollection alone.
  • Prefer this priority order when choosing what to continue:
    1. explicit user priority for the current turn
    2. previous round's explicit Next
    3. repo-native current-slice evidence in README, roadmap, tasklist, progress docs, or validation scripts
    4. the active mainline ladder already established for the project
  • Verification is mandatory. At minimum include:
    • one build / compile / typecheck style check
    • one smallest behavior check tied directly to the slice
  • For UI, control-flow, or recovery work, prefer a deterministic micro-check over a vague manual statement. If the repo has no test harness, add the smallest reproducible verification path that proves the slice.
  • Optimization is local cleanup for the completed slice only. Do not smuggle in adjacent scope.
  • Planning must recommend exactly one next slice.
  • If a delegated coding agent stalls, first shrink the slice and retry once if realistic; otherwise close the round manually, record the blocker, and still emit the next-round handoff.
  • Use codegraph or similar tools only to narrow the change surface or confirm call chains, never as a substitute for implementation or verification.

Workflow

  1. Identify the current mainline. Read the repo state, recent validation evidence, and any explicit Next handoff from the previous round.

  2. Choose one slice. Define the smallest meaningful increment that advances the current mainline. Refuse implicit multi-slice expansion.

  3. Develop the slice. Make the code changes needed to complete only that increment.

  4. Verify the slice. Collect concrete evidence. Prefer repo-native commands and tests. Required minimum:

    • one compile/build/typecheck-style check
    • one minimal behavior check tied to the slice
  5. Optimize the slice. Tighten naming, edge handling, comments, or small structure only where it directly improves the completed slice.

  6. Plan the next round. End with exactly these four sections:

    • Done: what changed this round
    • Verified: exact checks run and what they proved
    • Risks: remaining uncertainty, gaps, or blockers
    • Next: one recommended next slice only

Validation Ladder

Use the smallest proof that still makes the slice credible.

  • For package or infra changes: build/typecheck plus one command-level smoke check.
  • For state-transition logic: add one deterministic script or focused test that exercises the failure mode directly.
  • For UI slices: prefer a narrow state/logic check first, then deploy-style verification if the repo already has that path.
  • When a verification helper is missing, adding one is allowed if it stays inside the same slice and directly proves the change.

Recovery Ladder

When a round starts wobbling, recover in this order:

  1. shrink the slice once
  2. retry with the narrower target
  3. if still blocked, close the round manually with explicit blocker evidence
  4. emit one clean Next slice that resumes from the last verified state

Do not let a stuck subagent or flaky tool erase the loop.

Mainline Priority Example

When the user does not provide a new priority, continue the existing mainline before opening a fresh track.

Example ladder:

  • remote control capability gaps
  • operator console visibility and feedback
  • safety confirmations and failure recovery
  • recording and replay scaffold

Keep this as a priority ladder, not a license to touch multiple items in one round.

Output Contract

Use a compact round summary in this shape:

Completed

  • concise description of the finished slice

Verified

  • exact command or check
  • exact command or check

Optimized

  • any local tightening applied after verification

Next

  • one slice only

If the user prefers the four planning exits exactly, preserve Done / Verified / Risks / Next in the final planning block.

Anti-Patterns

  • Do not treat a broad milestone as a slice.
  • Do not skip verification because the change looks obvious.
  • Do not bury multiple future options under Next.
  • Do not rely on chat history as the source of truth when the repository says otherwise.
  • Do not use the optimization step to start the next feature.
  • Do not call a round verified if only the broad repo build passed but the slice-specific failure mode was never checked.
安全使用建议
Install this if you want Codex to run coding work in small verified rounds. It may guide the agent to inspect repo files, edit project code, and run tests or build checks when you ask for development work, but it does not add hidden execution, special privileges, or persistence.
能力评估
Purpose & Capability
The skill’s instructions match its stated purpose: structure coding work into development, verification, optimization, and next-step planning rounds.
Instruction Scope
Instructions are bounded to one slice per round, require repo evidence and verification, and discourage scope expansion; file edits and test commands are expected for the coding workflow.
Install Mechanism
The inspected artifact contains only SKILL.md markdown, with no executable scripts, dependencies, install hooks, bundled binaries, or obfuscated content.
Credentials
Reading repository docs/state and running build, typecheck, or focused behavior checks is proportionate to the stated development-loop purpose.
Persistence & Privilege
No background persistence, privilege escalation, credential/session access, destructive automation, or exfiltration behavior is described.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install continuous-dev-loop
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /continuous-dev-loop 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.0
V2: stronger verification, recovery ladder, and explicit Next handoff
v1.0.0
Initial release
元数据
Slug continuous-dev-loop
版本 1.1.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 2
常见问题

Continuous Dev Loop 是什么?

单-slice 持续开发循环 V2:更强验证、抗卡顿恢复、明确 Next 接力。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 47 次。

如何安装 Continuous Dev Loop?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install continuous-dev-loop」即可一键安装,无需额外配置。

Continuous Dev Loop 是免费的吗?

是的,Continuous Dev Loop 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Continuous Dev Loop 支持哪些平台?

Continuous Dev Loop 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Continuous Dev Loop?

由 ModeYapu(@modeyapu)开发并维护,当前版本 v1.1.0。

💬 留言讨论