技能详情（站内镜像，无评论）
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v1.0.0
统计：⭐ 0 · 24 · 0 current installs · 0 all-time installs
⭐ 0
安装量（当前） 0
🛡 VirusTotal ：良性 · OpenClaw ：良性
Package：bwiley1989/safe-long-run-mode-gpt54-claude
安全扫描（ClawHub）
- VirusTotal ：良性
- OpenClaw ：良性
OpenClaw 评估
This is an instruction-only guidance skill for running long tasks across GPT-5.4 and Claude; its requirements and actions are coherent with the stated purpose and it does not request extra credentials or install anything.
目的
The skill is purely procedural guidance (phases, checkpoints, model routing) and matches the name/description. It references external systems (Azure, Graph, Orgo, browser automation) only as contexts where long-run practices apply; it does not require unrelated binaries, env vars, or installs.
说明范围
The instructions tell the agent to write artifacts to disk, use subagents, validate auth, and interact with external APIs/GUI automation. Those operations are appropriate for long-running workflows, but they are broad: follow-on actions (auth/access to APIs, saving screenshots, writing files) will depend on the agent/platform connectors. The skill itself does not supply code or explicit endpoints.
安装机制
No install spec and no code files — lowest-risk pattern. Nothing will be written to disk by an installer and there are no download URLs to evaluate.
证书
The skill declares no required environment variables or credentials, which is proportionate. However, it prescribes interactions with cloud control planes and external APIs; those will require platform-supplied credentials at runtime. Users should ensure any credentials granted to the agent are minimal-scope and appropriate.
持久
Defaults are used (always:false, agent-invocable:true). The skill does not request permanent presence, nor does it modify other skills' configurations. Autonomous invocation is allowed by platform default and is not a concern on its own.
综合结论
This skill is a set of best-practice instructions — it does not install software or ask for secrets, so it's internally coherent. Before enabling it: (1) confirm where the agent will save checkpoints and ensure those storage locations are secure and have appropriate access controls; (2) restrict any cloud/API credentials the agent uses to least privilege and review auditing/expiration policies; (3) ensure subagent/autonomous runs are acceptabl…
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「Safe Long-Run Mode GPT5.4 + Claude」。简介：Operate long-running AI tasks safely across GPT-5.4 and Claude by using model s…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/bwiley1989/safe-long-run-mode-gpt54-claude/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
name: safe-long-run-mode-gpt54-claude
description: Operate long-running AI tasks safely across GPT-5.4 and Claude by using model selection rules, phased execution, checkpoints, resumable workflows, API throttling discipline, and subagent isolation. Use when a task may run for a while, touch multiple files/systems, involve external APIs, browser automation, Azure, Orgo, or multiple subagents, or when the user asks about long autonomous runs, rate limits, reliability, or safe operating mode.
---
# Safe Long-Run Mode (GPT-5.4 + Claude)
Use this skill for tasks that may run long, span multiple systems, or risk losing progress if interrupted.
## Core rule
Do not run long tasks as one monolithic attempt. Split into phases, write checkpoints, and keep the work resumable.
## Model selection
Use **GPT-5.4** for:
- coding
- docs
- research
- file-heavy transformations
- multi-agent delegated work
- repetitive build tasks
- long internal work where cost and throughput matter
Use **Claude** for:
- strategic judgment
- sensitive decisions
- nuanced synthesis
- client-facing polish
- brand voice refinement
- high-trust orchestration
Default to **GPT-5.4 first**. Escalate to Claude only when the task actually benefits from higher-quality judgment or tone.
## Operating procedure
### 1. Scope before acting
Before starting, decide:
- what the final deliverable is
- which systems/tools will be touched
- what can fail or throttle
- what must be saved after each phase
### 2. Break work into phases
Use phases such as:
1. gather / inspect
2. plan / write brief
3. execute / edit / build
4. validate
5. deploy or report
At the end of each phase, write artifacts to disk.
### 3. Always checkpoint
For long tasks, save progress in files:
- draft outputs
- notes
- reports
- partial results
- tracker entries
- checkpoint summaries
Prefer a resumable workspace state over a perfect one-shot run.
### 4. Isolate long work
Use subagents when:
- the task will take more than a few tool calls
- multiple files/systems are involved
- external APIs are involved
- failure should not pollute the main session
- specialized work can be delegated cleanly
### 5. Throttle external systems
When interacting with Azure, Graph, Orgo, messaging providers, registries, websites, or any external API:
- batch reads when possible
- avoid tight polling loops
- serialize risky writes
- respect retry/backoff
- avoid one-item burst loops when a bulk operation is possible
### 6. Prefer resumability over perfection
The goal is not "never fail." The goal is: if interrupted, resume with minimal loss.
## System-specific guidance
### Azure / cloud control planes
- validate auth first
- create foundational resources first
- verify after each layer
- log resource names/IDs
- do not chain long destructive commands blindly
### Browser / Orgo / GUI automation
- use explicit goals and stop conditions
- capture screenshots at checkpoints
- bound retry counts
- save artifacts locally
- prefer API/CLI over GUI when equivalent exists
### Coding / documentation work
- create a brief/spec first for complex tasks
- write files in chunks
- validate after each major change
- leave notes for resume if work is unfinished
## What to tell the user
When relevant, explain that safe long-run mode means:
- cheapest adequate model
- phased execution
- saved checkpoints
- subagent isolation
- controlled API usage
- resumable progress
## Failure handling
If a long task is interrupted:
1. summarize completed phases
2. point to saved artifacts
3. identify exact next step
4. resume from checkpoint rather than restarting
## References
- Read `references/checklist.md` for a reusable pre-flight checklist and model routing matrix.