Hi - I answer from the OpenSmartRoute documentation: routing, the API, plans and quotas, self-hosting. Ask away, or open a support ticket if you need a person.
Grounded in the docs - follow a source before acting on it.
self-improving-agent - Skill - OpenSmartRoute
Skillv1.0.0
self-improving-agent
Capture operational mistakes, repeated failures, and workflow lessons so future runs improve instead of repeating the same errors.
Imported from Wang200935/openclaw-skills (misc/misc/misc/misc/misc/misc/misc/ai/self-improving-agent/SKILL.md). Install upstream with npx skills add Wang200935/openclaw-skills --skill self-improving-agent. Copyright stays with the author.
Self-Improving Agent
Use this skill to turn mistakes into durable improvements.
Core workflow
Detect the failure, friction, or lesson.
Classify it:
command failure
bad assumption
workflow mistake
tool mismatch
communication mistake
Record the minimal durable lesson in the right place.
Promote repeated lessons into:
memory
AGENTS.md
TOOLS.md
a skill reference
Prefer specific correction rules over vague reflections.
Good lesson format
What happened
Why it failed
What to do differently next time
Do not
Write private secrets into broad shared notes
Create bloated postmortems for trivial mistakes
Record lessons without an action rule
Use it
Copy one of these into your project. Installing also returns the manifest and these snippets.
# after Install: the listing is in your workspace's routing pool - a plan picks it for its slot
curl -s -X POST https://api.opensmartroute.ai/api/v1/route -H 'Authorization: Bearer $OSR_API_KEY' -H 'Content-Type: application/json' -d '{"text": "...", "plan": true}'
Manifest
An Open Capability Manifest: the router reads it to know what this does, what it costs and when to pick it.
{
"ocm": "1",
"id": "wang200935-openclaw-skills-self-improving-agent",
"kind": "skill",
"name": "self-improving-agent",
"description": "Capture operational mistakes, repeated failures, and workflow lessons so future runs improve instead of repeating the same errors. Use when a task fails, a pattern keeps causing problems, a better method is discovered, or the agent should turn execution lessons into durable notes, checklists, or memory updates.",
"publisher": "Wang200935",
"version": "1.0.0",
"capabilities": {
"domains": [
"general"
],
"tags": [
"skill-md",
"github"
],
"languages": [
"en"
]
},
"quality_prior": 0.6,
"examples": [
"Capture operational mistakes, repeated failures, and workflow lessons so future runs improve instead of repeating the same errors. Use when a task fails, a pattern keeps causing problems, a better method is discovered, or the agent should turn execution lessons into durable notes, checklists, or memory updates."
],
"primary": false,
"metadata": {
"source": {
"provider": "github",
"repository": "https://github.com/Wang200935/openclaw-skills",
"path": "misc/misc/misc/misc/misc/misc/misc/ai/self-improving-agent/SKILL.md",
"ref": "8f2b5b95360219519f8692b3056a02a12aa76487",
"url": "https://github.com/Wang200935/openclaw-skills/blob/8f2b5b95360219519f8692b3056a02a12aa76487/misc/misc/misc/misc/misc/misc/misc/ai/self-improving-agent/SKILL.md",
"key": "Wang200935/openclaw-skills/misc/misc/misc/misc/misc/misc/misc/ai/self-improving-agent/SKILL.md"
}
},
"instructions": "# Self-Improving Agent\n\nUse this skill to turn mistakes into durable improvements.\n\n## Core workflow\n\n1. Detect the failure, friction, or lesson.\n2. Classify it:\n - command failure\n - bad assumption\n - workflow mistake\n - tool mismatch\n - communication mistake\n3. Record the minimal durable lesson in the right place.\n4. Promote repeated lessons into:\n - memory\n - AGENTS.md\n - TOOLS.md\n - a skill reference\n5. Prefer specific correction rules over vague reflections.\n\n## Good lesson format\n\n- What happened\n- Why it failed\n- What to do differently next time\n\n## Do not\n\n- Write pri",
"cost": {
"context_tokens": 181
}
}
Fetch it by URL: GET /api/v1/registry/wang200935-openclaw-skills-self-improving-agent/manifest?version=1.0.0
Reviews
Star ratings from people who tried it. One review per account; edit yours any time.
No reviews yet. Install it, try it, and be the first to rate it.