Xpersona
O

Openclaw Agent Optimize

CLAWHUB · Unverified

Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace (cost, model routing, context discipline, delegation, r... Skill: Openclaw Agent Optimize Summary: Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace (cost, model routing, context discipline, delegation, r... Tags: latest:1.2.1 Version history: v1.2.1 | 2026-03-08T19:12:55.372Z | user Add ambient-skill reduction guidance, authoritative context metrics, and verify/rollback-first optimization workflow. v1.2.0 | 2026-03-01T11:0

62
Rank
84
Safety
14k
Downloads

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 13.9K downloads reported by the source. Last updated 11/16/2026.

Avoid when

  • Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Install and run

Setup complexity: low.

clawhub skill install unknown:openclaw-agent-optimize
  1. Setup complexity is classified as HIGH. You must provision dedicated cloud infrastructure or an isolated VM. Do not run this directly on your local workstation.
  2. Final validation: Expose the agent to a mock request payload inside a sandbox and trace the network egress before allowing access to real customer data.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-unknown-openclaw-agent-optimize/snapshot"

Run-check

$0.02 USD

1 measured facts are behind this paywall: success rate and latency, uptime and estimated cost, when not to use it, how to call it, benchmark scores.

Agents pay $0.02 in USDC. A card payment is $0.50, the smallest a card allows.

Documentation

CLAWHUB

149,894 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: openclaw-agent-optimize
slug: openclaw-agent-optimize
version: 1.2.1
license: MIT
description: |
  Use when: you want a structured audit -> options -> recommended plan to improve an OpenClaw workspace
  (cost, model routing, context discipline, delegation, reliability).
  Don't use when: you want immediate config/cron mutations without review, or the question is unrelated to OpenClaw ops.
  Output: a prioritized plan with exact change proposals, expected impact, and rollback steps. No persistent changes without explicit approval.
triggers:
  - optimize agent
  - optimizing agent
  - improve OpenClaw setup
  - agent best practices
  - OpenClaw optimization
metadata:
  openclaw:
    emoji: "🧰"
---

# OpenClaw Agent Optimization

Use this skill to tune an OpenClaw workspace for **cost-aware routing**, **parallel-first delegation**, and **lean context**.

## Default posture

This skill is **advisory first**.
It should produce:
- audit,
- options,
- recommended plan,
- exact patch proposal,
- rollback,
- verification plan.

No persistent mutations without explicit approval.

## Quick start

1) **Full audit (safe, no changes)**
> Audit my OpenClaw setup for cost, reliability, and context bloat. Output a prioritized plan with rollback notes. Do NOT apply changes.

2) **Context bloat / transcript noise**
> My OpenClaw context is bloating (slow replies / high cost / lots of transcript noise). Identify the top offenders (tools, crons, bootstrap files, skills) and propose the smallest reversible fixes first. Do NOT apply changes.

3) **Model routing / delegation posture**
> Propose a model routing plan for (a) coding/engineering, (b) short notifications/reminders, (c) reasoning-heavy research/writing. Include an exact config patch + rollback plan, but do NOT apply changes.

## What good output looks like

- **Executive summary**
- **Top drivers**
  - cost
  - context
  - reliability
  - operator friction
- **Options A/B/C** with tradeoffs
- **Recommended plan** (smallest safe change first)
- **Exact proposals** + **rollback** + **verify**

## Safety contract

- Do not mutate persistent settings without explicit approval.
- Do not create/update/remove cron jobs without explicit approval.
- If an optimization reduces monitoring coverage, present options and require choice.
- Before any approved change, show:
  1. exact change,
  2. expected impact,
  3. rollback plan,
  4. post-change verification.

## High-ROI optimization levers

### 1) Output discipline for automation
Make maintenance loops truly silent on success.

### 2) Separate work from notification
If you want alerts but want interactive context lean:
- do the work quietly
- notify out-of-band with a short human receipt

### 3) Bootstrap discipline
Keep always-injected files short and load-bearing only.
Move long runbooks into `references/` or adjacent notes.

### 4) Ambient specialist surface reduction
A common hidden tax is **too many always-visible specialist skills**.
If a workflow is 

README.md

# openclaw-agent-optimize

> OpenClaw agent optimization skill — model routing, context management, delegation, and cron best practices.

An installable [OpenClaw](https://openclaw.ai) skill that packages battle-tested agent-optimization patterns. Drop it into your workspace and get instant guidance on cost-aware model routing, parallel-first delegation, lean context management, and more.

## 🙏 Credits & Inspiration

This skill is heavily inspired by **[affaan-m/everything-claude-code](https://github.com/affaan-m/everything-claude-code)** — a production-ready collection of Claude Code configs evolved over 10+ months by an Anthropic hackathon winner. The core philosophy of tiered model selection, progressive disclosure, parallel orchestration, and continuous learning was extracted and adapted from that work for the OpenClaw ecosystem.

Thank you [@affaan-m](https://github.com/affaan-m) 🎉

## 📦 What's Inside

| File | Purpose |
|------|---------|
| `SKILL.md` | Skill entry-point — triggers + quick-start workflow |
| `references/model-selection.md` | Tiered model routing guide |
| `references/context-management.md` | Context window discipline + progressive disclosure |
| `references/agent-orchestration.md` | Parallel-first delegation, split-role sub-agents |
| `references/cron-optimization.md` | Cron job patterns and model assignment |
| `references/memory-patterns.md` | Daily + long-term memory file design |
| `references/continuous-learning.md` | Hooks → instincts → skills pipeline |
| `references/safeguards.md` | Anti-loop + budget guardrails |

## 🚀 Install

### Via ClawHub (recommended)
[<img src="https://clawhub.ai/badge.svg" height="24">](https://clawhub.ai/phenomenoner/openclaw-agent-optimize)

```bash
npx clawhub install openclaw-agent-optimize
```

If you don’t see the skill immediately after a publish/update, it may be temporarily hidden while a security scan is running. Try again in a few minutes.

**Canonical page:** https://clawhub.ai/phenomenoner/openclaw-agent-optimize

### Manual
Copy the folder into your OpenClaw skills directory.

Common defaults (may differ by installation):
- macOS/Linux: `~/.openclaw/workspace/skills/`
- Windows: `%USERPROFILE%\.openclaw\workspace\skills\`

(If you’re unsure where your install lives, search for a `skills/` folder containing `openclaw-agent-optimize/`.)

## 📖 How to Use

The skill auto-triggers when you ask about optimizing your agent, improving your OpenClaw setup, or following agent best practices.

Key guidance you’ll get:
- **Native heartbeat is expensive** (it can load large main-session context) and **isn’t always reliably isolatable** in real deployments.
- Recommended strategy: disable native heartbeat and use an isolated heartbeat cron (alert-only).
- Persistent actions are **user-gated**: no config/cron mutations without explicit approval.
- Bonus: pair isolated heartbeat with **openclaw-mem** for cheap “RAG-style” task awareness.

## 📄 License

MIT

_meta.json

{
  "ownerId": "kn78kb299hg33qbk0pt7ffj8ks80hghm",
  "slug": "openclaw-agent-optimize",
  "version": "1.2.1",
  "publishedAt": 1772997175372
}

references/agent-orchestration.md

# Agent Orchestration (Parallel-First + Cost-Aware)

## Core rules
- **Parallelize independent tasks.** If task B does not depend on task A, spawn both.
- Keep main session clean: delegate long/noisy work to isolated sub-agents.

## Cost-aware delegation
- Pass **minimal context** (goal, constraints, file paths).
- Prefer sub-agents for:
  - long-running research
  - multi-step automation
  - work that benefits from isolation (cron-like behavior)

## Output discipline
- Sub-agents should:
  - write artifacts to files
  - return a short summary + next actions
  - avoid dumping long tool outputs into chat

## Delivery discipline
- For automated runs, prefer **alert-only** delivery:
  - `NO_REPLY` if nothing changed
  - message only for new actionable signals or errors

## Split-role delegation (when needed)
- Engineer: implement
- Reviewer: correctness & style
- Security: risk assessment

## Logging / recovery
- Log sub-agent runs to `memory/delegation_log.json` for crash recovery.
- Rotate/trim logs if they grow without bound.

references/context-management.md

# Context Management (Progressive Disclosure + Cost Hygiene)

## Window discipline
- Don’t enter the last ~20% of context when doing multi-file edits or refactors.
- If near the limit, prefer `/compact` **before** continuing.

## Progressive disclosure (3 levels)
1. **Metadata always visible:** filenames, brief summaries, TODOs.
2. **Body on trigger:** load only the file/section needed.
3. **Heavy resources on demand:** move schemas/docs to `references/` or scripts.

## Cost hygiene patterns
- Avoid repeatedly pasting large tool outputs into the conversation.
- Prefer scripts that output a small, bounded summary.
- Use fixed-size summaries: top K items, max X lines, max Y chars.

## OpenClaw practices
- Store large/static content under `references/`.
- Store state in small JSON files; rotate/trim logs that grow without bound.
- If session store accumulates stale cron sessions/orphan transcripts, run `context-clean-up` session-store hygiene (dry-run first, backup-first apply).
- Batch related decisions into a single user question (reduce back-and-forth).

Sponsored

Ads related to Openclaw Agent Optimize and adjacent AI workflows.