Adapt

Skill Walliiee/agent-harness/plugin/skills/adapt

A drop-in self-improving memory + operations harness for AI agents (OpenClaw / Hermes / Claude Code / Codex). Layered memory, autonomous drift-fixing loop, evals, and one-command DR. Templates + scripts, no secrets.

Install

npx -y skills add Walliiee/agent-harness --skill adapt

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Wire the agent-harness templates into the current project by driving scripts/adapt.py — probe the machine, render the templates from the user's home/org/agents, validate, show a diff, and apply only after explicit confirmation. Invoke when the user wants to install or set up agent-harness.

SKILL.md

2.5 KB, 561 tokens by cl100k_base, as published. Nobody here has run it

Adapt the agent-harness to this project

You are wiring the agent-harness templates into the user's setup using the bundled scripts/adapt.py. The plugin's files are under ${CLAUDE_PLUGIN_ROOT}.

Hard rule: never write to a real home directory without an explicit, separate confirmation from the user. adapt.py defaults to a dry-run that writes nothing — keep it that way until the user has seen the diff and said go.

Inputs

Parse $ARGUMENTS for these, and ask for any that are missing (do not guess):

--home — the harness home dir (commonly ~/.openclaw).
--gh-org — the GitHub org/owner for the per-workspace remotes.
--agents — comma-separated roster (default main,dev if the user has no preference).
--telegram-chat-id — optional notification channel id; omit if not given.

Steps

Preflight first. Run the clone-level check and summarize anything missing:
```
bash "${CLAUDE_PLUGIN_ROOT}/scripts/preflight.sh"
```
If a REQUIRED dependency is missing, stop and tell the user — do not proceed to apply.

Probe the machine so the user sees what was auto-detected:

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/adapt.py" --probe-only

Dry-run (writes nothing) with the resolved inputs, and show the user the staged diff verbatim:

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/adapt.py" \
  --home <home> --gh-org <org> --agents <roster>

Confirm. Ask the user to review the diff and explicitly approve writing. Do not continue without a clear yes.

Apply only after approval:

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/adapt.py" \
  --home <home> --gh-org <org> --agents <roster> --apply --out <home>

Verify: run bash "${CLAUDE_PLUGIN_ROOT}/scripts/scrub-audit.sh" and report the result. Point the user at docs/getting-started.md for installing the schedulers and the live-stack steps.