一键在 Manus 中运行任何 Skill

开始使用

verification-before-completion

星标3

分支0

更新时间2026年7月10日 16:35

Use when about to claim work complete, fixed, passing, or ready, including before moving to the next planned task.

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

aultramen

aultramen/super-compound

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Verification Before Completion

Overview

Evidence precedes every success claim. Confidence, a changed diff, and another agent's report are not evidence.

When to Use

Apply this gate before any completion or satisfaction wording, task transition, commit, push, PR claim, or delegation that assumes preceding work succeeded.

For a test, build, lint, bug-fix, or delegated-work claim, read Claim Evidence.
For a completed feature, workflow, plan, or gap-closure outcome, also read Goal-Backward Verification.
For work spanning components, services, layers, auth, configuration, events, or background jobs, read integration-checking, run its checks, then return to this gate. Detailed cross-component checks live there, not here.

Completion Gate

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

Identify the command or inspection that proves the exact claim.
Run it now, fully and freshly.
Read all relevant output, the exit code, and failure counts.
Compare the evidence with the claim and requirements.
If it fails or is incomplete, report actual status and gaps. Otherwise cite the evidence with the claim.

No executable proof available? Use an explicit requirement checklist and direct artifact inspection; disclose residual uncertainty. Never convert missing evidence into success wording.

Red Flags

Thought	Required response
"It should work" / "I'm confident"	Run the proving check.
"A partial suite is enough"	Run the full relevant check or narrow the claim.
"The subagent said it passed"	Verify independently.
"The linter passed, so the build/tests pass"	Each claim needs its own evidence.
"Different wording avoids the rule"	Implication of success is still a claim.

Integration

executing-plans invokes this before task completion.
test-driven-development supplies red-green evidence.
systematic-debugging supplies symptom and regression evidence.
code-review verifies findings before reporting them.
gap-closure re-runs the original verification.
state-management records verified status only after this gate.

同仓库更多 Skills

同仓库

agentic-delivery

aultramen/super-compound

Use when following the Super Compound BRD -> PRD -> FSD -> GOAL -> IMPLEMENTATION -> VERIFICATION delivery path, artifact traceability, FSD authority, optional ADR handling, zero context bloat issue slicing, or OPEN-* stop conditions.

2026-07-163

executing-plans

aultramen/super-compound

Use when an approved FSD goal or lightweight issue pointer is ready for implementation.

2026-07-163

integration-checking

aultramen/super-compound

Use when multi-component work needs cross-component verification of API contracts, data flow, events, auth, config, UI, background jobs, or tests.

2026-07-163

issue-workflow

aultramen/super-compound

Use when /sc-plan needs FSD GOAL-* packets turned into lightweight issue pointers, local Markdown Kanban boards, blocker DAGs, or multi-agent task contracts.

2026-07-163

parallel-execution

aultramen/super-compound

Use when a plan or issue board has 2+ independent execution streams whose time saving exceeds coordination overhead. Dispatches agents in isolated git worktrees only after required delivery gates pass.

2026-07-163

plan-verification

aultramen/super-compound

Use when an FSD and goal issue board need requirement coverage, goal quality, dependency DAG, sizing, and verification validated before execution.

2026-07-163

name	verification-before-completion
description	Use when about to claim work complete, fixed, passing, or ready, including before moving to the next planned task.

Verification Before Completion

Overview

Evidence precedes every success claim. Confidence, a changed diff, and another agent's report are not evidence.

When to Use

Apply this gate before any completion or satisfaction wording, task transition, commit, push, PR claim, or delegation that assumes preceding work succeeded.

For a test, build, lint, bug-fix, or delegated-work claim, read Claim Evidence.
For a completed feature, workflow, plan, or gap-closure outcome, also read Goal-Backward Verification.
For work spanning components, services, layers, auth, configuration, events, or background jobs, read integration-checking, run its checks, then return to this gate. Detailed cross-component checks live there, not here.

Completion Gate

NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE

Identify the command or inspection that proves the exact claim.
Run it now, fully and freshly.
Read all relevant output, the exit code, and failure counts.
Compare the evidence with the claim and requirements.
If it fails or is incomplete, report actual status and gaps. Otherwise cite the evidence with the claim.

No executable proof available? Use an explicit requirement checklist and direct artifact inspection; disclose residual uncertainty. Never convert missing evidence into success wording.

Red Flags

Thought	Required response
"It should work" / "I'm confident"	Run the proving check.
"A partial suite is enough"	Run the full relevant check or narrow the claim.
"The subagent said it passed"	Verify independently.
"The linter passed, so the build/tests pass"	Each claim needs its own evidence.
"Different wording avoids the rule"	Implication of success is still a claim.

Integration

executing-plans invokes this before task completion.
test-driven-development supplies red-green evidence.
systematic-debugging supplies symptom and regression evidence.
code-review verifies findings before reporting them.
gap-closure re-runs the original verification.
state-management records verified status only after this gate.