decision-gates SKILL.md skillA

decision-gates from ooooooooooooooooooop/agent-tools is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add ooooooooooooooooooop/agent-tools/decision-gates@v0.2.0

Pins v0.2.0 in mdr.lock. Pin the fingerprint instead.

Audit grade
A, 17 of 17 checks passed
Kind
SKILL.md skill
Repository
ooooooooooooooooooop/agent-tools
Path
skills/decision-gates/SKILL.md
Versions recorded
2
Latest version
v0.2.0, committed Aug 28, 2026
First version recorded
Aug 23, 2026
Size
8,650 bytes
Repository stars
4
License
MIT

在 DSH 长程任务中为决策层加装五道防偏闸门:原始证据锚、对抗式对齐审计、跨包一致性校验、成本比对、防御自检,防止目标漂移、验收失真与摘要偏差。用于跨会话长任务、多 worker 并行、审计类任务、成本敏感任务的 checkpoint 记录与恢复,以及任何需要"决策层不跑偏"保障的执行场景。.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# 决策闸门(Decision Gates)

## 核心规则

决策层跑偏的根因不是提示词不够强,而是**状态没有物化**:目标只存在于对话里、验收只依赖自述、恢复只信任摘要。决策闸门把防偏从"意志"变成"硬门禁":

1. **闸门 0 原始证据锚**:checkpoint 必须同时落盘摘要与原始证据锚 JSON;恢复时先读锚、后读摘要,锚里有 SKIP/WARN/PENDING 则"PASS"自动降级为"PARTIAL"。
2. **闸门 1 对抗式对齐审计**:**事件驱动**触发一个**不读摘要**的独立审计 subagent,只给 objective + 原始证据锚 + 产出物,防"自检即跑偏"。触发事件:objective 变更 / 关键假设被证伪 / 不可逆或高风险边界 / 多 worker 结果集成 / 证据相互冲突 / 连续无 Progress Delta / 重大架构决策 / **最终验收(必跑)**。若连续 ≥5 个 checkpoint 无任何触发事件,补一次低成本自检兜底。固定周期计数不再是主要触发方式(防止无行动价值的重复审计)。
3. **闸门 2 跨包一致性校验**:≥2 个 worker 并行时,验收前先跑算法脚本核对文件覆盖重叠/空白/哈希冲突,0 token。
4. **闸门 3 成本比对**:每个派工单记录授权档位与预算,worker 返回时核对实际档位与消耗,0 token。
5. **闸门 4 防御自检**:每 5 个 checkpoint 扫描历史锚,统计摘要比证据乐观的次数(summary_bias),≥2 次判定 DEFENSE_DRIFT 并回退,0 token。
6. **闸门 5 物理范围锁(Scope Lock)**:验收或交付前用 `gate_scope_lock.py` 严格校验变更文件集,任何触碰 Non-Goals 或越界文件直接判 `SCOPE_VIOLATION`,0 token。
7. **闸门 6 轨迹死循环熔断(Deadband Loop Interceptor)**:连续 2 次相同报错/失败模式时通过 `gate_deadband.py` 自动触发 `DEADBAND_TRIPPED`,强制推翻假设并回退,0 token。

**零 token 原则**:闸门 2/3/4/5/6 是纯数据算法逻辑,必须用 `scripts/` 下的 Python 脚本执行,禁止用 LLM 调用替代;这些机械 scope/cost/integrity 检查保持低成本执行,不因事件化而删除。唯一消耗 token 的闸门 1 用最便宜档位的独立 subagent,按上方事件驱动触发。

**验证预算(VERIFICATION_PURPOSE)**:任何验证/审计动作执行前必须能回答:
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add ooooooooooooooooooop/agent-tools/decision-gates@sha256:f4b7197e1ee06839

The line at the top of this page pins the label v0.2.0, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_jipld5zt53claa77.svg)](https://markdownregistry.com/a/art_jipld5zt53claa77)

3 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 2 versions of this file, between Aug 23, 2026 and Aug 28, 2026, about one every 5 days.

versioncommittedcommitsizeaudit
v0.2.0 latest2026-08-28 278776e 8,650 BA view · diff
v0.1.02026-08-23 8753690 7,288 BA view

Audit of the latest version

A  17 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (8650 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No link to a raw IP address
  • pass: No script tag

Source

GitHub

ooooooooooooooooooop/agent-tools · 4 stars · license MIT · pushed 2026-10-23 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_jipld5zt53claa77
GET https://markdownregistry.com/api/v1/resolve?ref=ooooooooooooooooooop/agent-tools/decision-gates
GET https://markdownregistry.com/api/v1/blob/f4b7197e1ee06839b990b29453bb4ba974f713404e21304ae01430adb466ffa0

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from ooooooooooooooooooop/agent-tools

AGENTS.md AGENTS.md file
ooooooooooooooooooop/agent-tools · AGENTS.md
git:20261022.bc4f1e5 · audit A · 4 stars
your-skill-name SKILL.md skill
ooooooooooooooooooop/agent-tools · _template/SKILL.md · 用一句中文说明这个 Skill 能做什么,以及何时应该触发它
v0.1.0 · audit A · 4 stars
AGENTS.md@dsh-config AGENTS.md file
ooooooooooooooooooop/agent-tools · dsh-config/AGENTS.md
git:20260914.0c9411e · audit A · 4 stars
llms.txt@research/agent-harness-survey-v1/changelogs/amp llms.txt file
ooooooooooooooooooop/agent-tools · research/agent-harness-survey-v1/changelogs/amp/llms.txt
git:20260921.a69e78d · audit A · 4 stars
llms.txt@research/agent-harness-survey-v1/changelogs/qoder llms.txt file
ooooooooooooooooooop/agent-tools · research/agent-harness-survey-v1/changelogs/qoder/llms.txt
git:20260921.a69e78d · audit A · 4 stars
llms.txt@research/agent-harness-survey-v1/changelogs/replit llms.txt file
ooooooooooooooooooop/agent-tools · research/agent-harness-survey-v1/changelogs/replit/llms.txt
git:20260921.a69e78d · audit A · 4 stars
chatgpt-web-bridge SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/chatgpt-web-bridge/SKILL.md · 通过 chatgpt-web 的 MCP 或 REST 桥,把本地 agent 接入已登录的 ChatGPT Web。 适用于“推进/继续网页端会话”“让网页端设计、本地执行”“协作/对等推进”,以及只读拉取网页端会话/项目资料和…
git:20260921.679541d · audit A · 4 stars
clarify-before-change SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/clarify-before-change/SKILL.md · 在修改前澄清范围、风险、假设与验收标准,避免误改。用于修改 Skill、AGENTS.md、项目规则、工作流规则、仓库结构、复杂代码或具有结构影响的文档,也用于优化 Skill、调整规则、重构、重组或需求不完整的场景。
git:20260828.278776e · audit A · 4 stars
dsh-config-sync SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/dsh-config-sync/SKILL.md · 把 DSH(DeepSeek Harness)的用户配置(~/.dsh 下的 AGENTS.md、settings.yaml 及可选 profiles/预设)打包成脱敏骨架并在设备间安全同步恢复,提供只读审计、显式应用与 SHA-256…
v1.0.0 · audit A · 4 stars
environment-bootstrap SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/environment-bootstrap/SKILL.md · 从源仓库向其他设备或用户级 Skill 目录恢复已登记的 Codex Skill,提供只读审计、显式应用、SHA-256 校验,并且不删除目标端额外文件。用于复现 Codex 工作环境、比较已安装 Skill…
git:20260817.d6c4635 · audit A · 4 stars
evolution-proposal SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/evolution-proposal/SKILL.md · 把 evolution-inbox 中的审计异常条目转化为可评审的进化提案(规则/技能/插件/路由补丁)。先做系统化问题分析(量化基线→全量聚合→根因分类→优先级排序),再用便宜模型起草、前沿模型评审、人工批准后走既有门禁固化。用于处理…
v0.3.0 · audit A · 4 stars
execution-discipline SKILL.md skill
ooooooooooooooooooop/agent-tools · skills/execution-discipline/SKILL.md · 在 DSH 长程/多工具会话中强制九条执行纪律,防止高频反模式重演:禁止无 wait 轮询、禁止把门禁拒绝当交付物抛回决策、外部研究先探测通道再极窄探针、探查派子代理不肉身…
v0.3.0 · audit A · 4 stars

Every file in ooooooooooooooooooop/agent-tools

Other files named decision-gates

decision-gates SKILL.md skill
uchicago-dsi/ai-sci-skills · skills/decision-gates/SKILL.md · Separate a scientific gate that decides promotion, continuation, stopping, or retirement from a mechanical check you…
git:20260921.9b3de7c · audit A · 20 stars
decision-gates SKILL.md skill
ooooooooooooooooooop/personal-ai · skills/decision-gates/SKILL.md · 在 DSH 长程任务中为决策层加装五道防偏闸门:原始证据锚、对抗式对齐审计、跨包一致性校验、成本比对、防御自检,防止目标漂移、验收失真与摘要偏差。用于跨会话长任务、多 worker 并行、审计类任务、成本敏感任务的 checkpoint…
v0.2.0 · audit A · 4 stars

Browse by kind, by grade A, or by owner.