Skip to content
Get packs

Guide · Agent skill

How to use the SkillOpt-Sleep OpenClaw Reference agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

skillopt-sleep is an agent skill that gives an AI agent context about a contributed reference adaptation of Microsoft's SkillOpt-Sleep for OpenClaw. The adaptation connects the shared SkillOpt-Sleep cycle to a DeepSeek chat API and sample task files from one contributor's machine. This guide explains what the skill does, how to install it, how to ask for help with it, and why you should not run its scripts unchanged.

SkillOpt-Sleep OpenClaw Referencemicrosoft/SkillOpt · MIT

Quick start

  1. 1Download the skill ZIP. Get the skillopt-sleep ZIP from this page. Keep a copy of the original so you can compare later edits.
  2. 2Install the skill. Unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder. Alternatively, upload it under Settings, Capabilities, Skills in the Claude apps.
  3. 3Read the README and SKILL.md. Start with README.md, which the source calls the authoritative status and adaptation guide. Confirm you understand why the files should not run unchanged.
  4. 4Ask a focused question. Request an explanation or review of one file, such as the DeepSeek backend, so the agent works on a bounded task.
  5. 5Test with the mock backend. Run python -m skillopt_sleep --backend mock to exercise the supported shared engine without a remote API.
  6. 6Plan the porting gaps. Work through the source's list: replace absolute paths, update the get_backend contract, adopt the current staging manifest, and add isolated tests.

What an agent skill is and what this one does

An agent skill is a folder of instructions and optional files that an AI agent loads when a request matches the skill's description. The skill's SKILL.md file holds that description and the guidance the agent follows.

This skill is a reference, not a finished tool. Its description says it is for studying or porting the contributed DeepSeek wrapper, not for use as a ready-to-run installation. Once installed, the agent can help you read, explain or port the code. It does not gain a new ready-to-use ability such as a working nightly optimizer.

  • Explains how the OpenClaw adaptation is structured and which files do what
  • Helps identify hardcoded paths, credentials and assumptions that need replacing
  • Helps plan a port of the custom backend to the current shared skillopt_sleep interfaces

Install the skillopt-sleep skill

Download the ZIP of the skill from this page. You can install it for your user account or for a single project.

For a user-wide install, unzip the archive so the skill folder sits inside ~/.claude/skills/. For a project-only install, place it inside the project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills.

  • User-wide: unzip into ~/.claude/skills/
  • Single project: unzip into the project's .claude/skills/ folder
  • Claude apps: Settings, then Capabilities, then Skills, then upload the ZIP

Ask for help with natural requests

You do not need a special command. Describe what you want in plain language and mention the OpenClaw SkillOpt-Sleep reference or the DeepSeek wrapper so the agent matches the skill.

Good requests ask for explanation, review or porting help. Avoid asking the agent to run the cron script or run_sleep.py as they are. The source says those files point at one user's home directory and can send task data to a remote API.

  • Explain how skillopt_sleep_openclaw.py calls the DeepSeek Chat Completions API
  • List the hardcoded absolute paths in run_sleep.py and config.json that I need to replace
  • Compare this custom backend with the shared skillopt_sleep CLI backends
  • Outline the steps to port the DeepSeek wrapper to the current get_backend call contract

Know what you need before you go further

Running or porting this code needs Python 3 and a clone of the SkillOpt repository, because the scripts import skillopt_sleep. You also need a DeepSeek API key set as DEEPSEEK_API_KEY, either in the environment or in ~/.openclaw/.env.

An Ollama service with nomic-embed-text is optional. The embedding helper is not wired into the cycle. Scheduling with the shell script needs Bash, cron, and curl and jq. Setup requires editing Python, shell and config files, so this is not a task for non-developers.

  • Python 3 and a clone of the SkillOpt repository
  • DEEPSEEK_API_KEY in the environment or ~/.openclaw/.env
  • Optional: local Ollama with nomic-embed-text
  • For shell scheduling: Bash, cron, curl and jq
  • An OpenClaw workspace layout that matches the hardcoded paths, or edits to replace them

Test SkillOpt-Sleep safely with the shared CLI

The source recommends the supported shared CLI for testing SkillOpt-Sleep itself until the OpenClaw wrapper's gaps are resolved. The mock backend does not call a remote API.

Treat this directory as source material for a future port, not as a way to run the nightly cycle.

  • Do not schedule the wrapper files without rewriting their paths and credentials
  • Shared scheduler actions manage shared-engine cron entries and do not run the custom backend
python -m skillopt_sleep --backend mock

Safety, data and credential boundaries

The DeepSeek backend sends task, skill, memory, response, rubric and reflection content to its configured Chat Completions endpoint. Do not assume this removes secrets or private details. Use synthetic or reviewed task files, replace sample names, URLs and machine paths, and load credentials through your secret-management setup.

Verify TLS and retention policy for every remote endpoint, and inspect all staged artifacts before adoption. The bundled fixtures are examples from one environment, so their scores and cost estimates do not show effectiveness or a stable price.

  • The skill runs scripts and uses the network
  • Remote endpoints receive task data, so review what each file sends
  • Config values are sample settings, not enforced runtime limits

What people use it for

  • Studying how a custom SkillOpt-Sleep backend is written against a DeepSeek Chat Completions API
  • Listing hardcoded paths and credentials that must be replaced before any real use
  • Planning a port of the DeepSeek wrapper into a supported OpenClaw integration
  • Comparing a custom backend with the shared skillopt_sleep CLI backends

Frequently asked questions

Is the SkillOpt-Sleep OpenClaw skill ready to run?

No. The skill's description and SKILL.md say it is a contributed reference, not a supported plug-and-play integration. The files contain environment-specific paths and parts have not been ported to current interfaces.

What do I need to install skillopt-sleep?

Install it by unzipping the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder, or by uploading it under Settings, Capabilities, Skills in the Claude apps. Running the code also needs Python 3, a SkillOpt repository clone and a DeepSeek API key.

Is skillopt-sleep free?

The source gives the licence as MIT. It does not list prices. The DeepSeek API is a separate service that you would need your own key for, and the source does not state its costs.

Why does the skill say not to run the cron script?

run_sleep_cron.sh is a machine-specific category runner, not a portable cron installer. It points at one user's home directory and assumes the contributor's workspace layout.

What are the alternatives?

For a supported setup, the source points to the shared python -m skillopt_sleep CLI, which offers backends such as mock, claude, codex, copilot, handoff and azure_openai. Use its shared documentation for scheduling and skill consolidation.

Does the skill send my data anywhere?

The DeepSeek backend sends task, skill, memory, response, rubric and reflection content to its configured Chat Completions endpoint. Use synthetic or reviewed task files and check the endpoint's TLS and retention policy first.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you are a developer studying how SkillOpt-Sleep backends are written, or a maintainer planning to port the DeepSeek wrapper into a supported OpenClaw integration. Pick the shared skillopt_sleep CLI instead if you want a working nightly optimizer or are not comfortable editing Python, shell and config files.

Download the skill