一键在 Manus 中运行任何 Skill

wf-auto

星标3

分支0

更新时间2026年7月4日 10:30

Perpetual auto-optimization mode. Never stops until 8-angle exhaustion. Inherits WF acceptance gates and subagent orchestration per cycle. Use for Claude /wf-auto, Codex $wf-auto, auto mode, or unbounded self-directed optimization.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

LiWeny16

LiWeny16/LetShare

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

WF Auto - Perpetual Auto-Optimization

Load

Harness/WF-AUTO.md
Harness/subagents.md
Harness/dispatch.md
Harness/agent-workflow.md
.claude/skills/wf-review/SKILL.md

Trigger

Claude /wf-auto
Codex $wf-auto
wf auto, auto mode, or a request for continuous self-directed improvement

Do not use when the user gives a bounded task, requests maximum parallelism (/wf-max), needs an urgent production hotfix, or the codebase is tiny enough that auto scanning costs more than it helps.

Hard Rules

Never stop except the A-GATE: all 8 angles empty, oracle empty, spark empty, and 2 confirmation rounds.
CEO never edits production source. CEO may write only Harness/tasks/auto/PLAN.md and Harness/tasks/auto/PROGRESS.md.
Dispatch all W0 sources in one batch when the runtime allows it: 8 angles, oracle, and spark searchers.
One accepted finding per cycle: <=3 files and <=50 changed lines. Larger ideas escalate to /wf or /wf-max, then return to auto.
Every accepted cycle inherits the full WF chain: Mini PRD -> AC IDs -> test/validation plan -> implementer -> verifier -> cross-review -> reflector PASS -> evidence ledger -> next W0.
Review is mandatory: spec review, code-quality review, then reflector.
Value Gate is scored for spark candidates: pass is >=18/25 and no dimension below 3.
Intent Checkpoint is adaptive: 2 -> 5 -> 10 cycles, exactly two questions.
Record compact evidence per cycle; do not paste full logs or transcripts.

Loop

W0: SENSE (8 angles + oracle + spark sources)
A-GATE: continue, oracle, spark, confirm, or stop
W1: PRIORITIZE one finding
W2-W5: Mini PRD -> AC -> test/validation plan -> implementer -> verifier -> cross-review -> reflector PASS
RECORD: evidence ledger
LOOP: next W0

Return

Report cycles run, findings addressed by source, evidence ledger, exhaustion evidence if any, weak spark count, final state, and residual risks.

同仓库更多 Skills

同仓库

subagent-orchestrator

LiWeny16/LetShare

Use when work needs bounded subagent coordination, parallel read-only exploration, independent review gates, broad context partitioning, or controlled handoffs.

2026-07-043

tdd

LiWeny16/LetShare

Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.

2026-07-043

wf-auto-spark

LiWeny16/LetShare

Perpetual inspiration mode for /wf-auto-spark or $wf-auto-spark. Inherits WF-AUTO/WF execution gates while using external spark search, roadmap anchoring, and <=50% deviation guard.

2026-07-043

wf-learn

LiWeny16/LetShare

Use for /wf-learn in Claude Code, $wf-learn or /skills wf-learn in Codex, or when the user wants a learning cycle from repeated errors, corrections, patterns, or session history.

2026-07-043

wf-max

LiWeny16/LetShare

Use for /wf-max in Claude Code, $wf-max or /skills wf-max in Codex, or maximum-parallelism Harness work with WF strict-superset gates and CEO -> manager -> worker decomposition.

2026-07-043

wf-readme

LiWeny16/LetShare

Use when a project README already exists and the user asks to preserve, merge, modernize, optimize, or clarify repository documentation during harness install or documentation work.

2026-07-043

name	wf-auto
description	Perpetual auto-optimization mode. Never stops until 8-angle exhaustion. Inherits WF acceptance gates and subagent orchestration per cycle. Use for Claude /wf-auto, Codex $wf-auto, auto mode, or unbounded self-directed optimization.

WF Auto - Perpetual Auto-Optimization

Load

Harness/WF-AUTO.md
Harness/subagents.md
Harness/dispatch.md
Harness/agent-workflow.md
.claude/skills/wf-review/SKILL.md

Trigger

Claude /wf-auto
Codex $wf-auto
wf auto, auto mode, or a request for continuous self-directed improvement

Do not use when the user gives a bounded task, requests maximum parallelism (/wf-max), needs an urgent production hotfix, or the codebase is tiny enough that auto scanning costs more than it helps.

Hard Rules

Never stop except the A-GATE: all 8 angles empty, oracle empty, spark empty, and 2 confirmation rounds.
CEO never edits production source. CEO may write only Harness/tasks/auto/PLAN.md and Harness/tasks/auto/PROGRESS.md.
Dispatch all W0 sources in one batch when the runtime allows it: 8 angles, oracle, and spark searchers.
One accepted finding per cycle: <=3 files and <=50 changed lines. Larger ideas escalate to /wf or /wf-max, then return to auto.
Every accepted cycle inherits the full WF chain: Mini PRD -> AC IDs -> test/validation plan -> implementer -> verifier -> cross-review -> reflector PASS -> evidence ledger -> next W0.
Review is mandatory: spec review, code-quality review, then reflector.
Value Gate is scored for spark candidates: pass is >=18/25 and no dimension below 3.
Intent Checkpoint is adaptive: 2 -> 5 -> 10 cycles, exactly two questions.
Record compact evidence per cycle; do not paste full logs or transcripts.

Loop

W0: SENSE (8 angles + oracle + spark sources)
A-GATE: continue, oracle, spark, confirm, or stop
W1: PRIORITIZE one finding
W2-W5: Mini PRD -> AC -> test/validation plan -> implementer -> verifier -> cross-review -> reflector PASS
RECORD: evidence ledger
LOOP: next W0

Return

Report cycles run, findings addressed by source, evidence ledger, exhaustion evidence if any, weak spark count, final state, and residual risks.