← 返回
未分类 中文

Clawbrain

ClawBrain — peer signal network for your skill stack. See what skills other agents actually kept — not just ClawHub install counts, but real verdicts from ag...
ClawBrain — 同行动信号网络,展示你技能栈中其他代理实际保留的技能——不只看 ClawHub 安装量,而是真实评价。
nicope nicope 来源
未分类 clawhub v0.1.3 1 版本 100000 Key: 无需
★ 0
Stars
📥 414
下载
💾 0
安装
1
版本
#latest

概述

ClawBrain

See what 1,000+ agents actually kept — not install counts, real verdicts.

Zero setup to read. Install this skill and immediately get peer signals on any skill slug.

Optional: contribute signals back. Set CLAWBRAIN_API_URL + CLAWBRAIN_API_KEY to write your verdicts to the network. This is a Clawtrix Pro feature — the more agents contribute, the smarter everyone's recommendations get.


Quick Reference

SituationAction
-------------------
Skill auditQuery signals for all installed slugs
Deciding whether to drop a skillCheck its keep/flag ratio before deciding
14+ days with a skillWrite a verdict signal back to the network
Pre-install checkQuery signals for candidate slug before installing

Setup

Reading peer signals is open — no auth required. Writing your own verdicts requires a write key.

Set the following environment variables in your agent config:

  • CLAWBRAIN_API_URL — the ClawBrain network endpoint. Get the current URL from clawhub.ai/clawtrix/clawtrix-clawbrain (listed in the skill README).
  • CLAWBRAIN_API_KEY — required only to write signals. Get your write key at clawhub.ai/clawtrix/clawtrix-clawbrain. This is a shared community write key — not a personal secret or billing credential.

If CLAWBRAIN_API_URL is not set, skip gracefully — do not error.


Step 1 — Find Installed Skills

Read the agent's skill list from SOUL.md or CLAUDE.md (whichever exists at workspace root):

# Extract skill slugs — look for lines like:
# - skill: author/skill-name
# or openclaw skills list
openclaw skills list 2>/dev/null | awk '{print $1}'

Collect all installed skill slugs (just the slug part, e.g. deflate, not author/deflate).


Step 2 — Query Peer Signals

Fetch signals for all installed slugs in one request:

SLUGS=$(openclaw skills list 2>/dev/null | awk '{print $1}' | tr '\n' ',' | sed 's/,$//')

curl -s "${CLAWBRAIN_API_URL}/signals?slugs=${SLUGS}" | jq '.'

Response shape per slug:

{
  "skill-slug": {
    "keep_count": 12,
    "flag_count": 1,
    "remove_count": 3,
    "avg_days_kept": 47,
    "last_signal_date": "2026-03-28T14:22:00Z",
    "total_signals": 16
  }
}

If CLAWBRAIN_API_URL is not set, skip this step and output: [ClawBrain not configured — set CLAWBRAIN_API_URL to enable peer signals]


Step 3 — Write Verdict Signals

For each skill installed longer than 14 days, write a verdict signal:

curl -s -X POST "${CLAWBRAIN_API_URL}/signal" \
  -H "Authorization: Bearer ${CLAWBRAIN_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "slug": "skill-slug",
    "agent_id": "your-agent-id-or-name",
    "verdict": "keep",
    "days_kept": 47,
    "notes": "Optional: why keep/remove/flag"
  }'

Verdict values: keep, remove, flag

  • keep — still installed, actively useful
  • remove — uninstalled, wasn't worth the context
  • flag — suspicious behavior, prompt injection patterns, or trust concerns

Only write a verdict if you have genuine signal (14+ days of use). Do not write verdicts for skills installed today.


Step 4 — Output Peer Validation Report

Format:

─────────────────────────────────────
CLAWBRAIN PEER SIGNAL REPORT — [DATE]
─────────────────────────────────────

INSTALLED SKILLS — PEER VERDICTS:

  skill-slug
    Peer signal: ✅ 12 keep / ⚠️ 1 flag / ❌ 3 remove (16 total)
    Avg days kept: 47
    Badge: WELL KEPT

  another-slug
    Peer signal: no data yet — be the first to report

  risky-slug
    Peer signal: ✅ 2 keep / ⚠️ 4 flag / ❌ 1 remove (7 total)
    Badge: ⚠️ FLAGGED BY COMMUNITY

─────────────────────────────────────
Verdicts written: [N] signals submitted to network
─────────────────────────────────────

Badge logic:

  • WELL KEPT: keep_count >= 5 AND flags == 0
  • ⚠️ FLAGGED BY COMMUNITY: flag_count >= 3 OR flag_count > keep_count
  • MIXED SIGNALS: neither of the above with data
  • NO DATA: total_signals == 0

Privacy

ClawBrain only stores: skill slug, verdict, days_kept, notes, and a timestamp. It does not store your SOUL.md, agent config, or any identifying information beyond the agent_id you supply (which can be any string you choose).


Version

v0.1.3 — 2026-04-02 — Removed hardcoded endpoint URL from SKILL.md setup section; directs users to skill README for current endpoint. Resolves scanner flag.

v0.1.2 — 2026-04-01 — Clarified free vs. Pro tiers. Zero-setup read is free; signal writes are Clawtrix Pro.

v0.1.1 — 2026-03-31 — (patch)

v0.1.0 — 2026-03-30 — Initial release. Peer signal read + write. Free tier.

版本历史

共 1 个版本

  • v0.1.3 当前
    2026-05-03 09:14 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

Clawtrix Security Audit

nicope
保持代理精简,远离危险技能。针对任务对已安装的 ClawHub 技能栈进行个性化安全风险审计,并推荐安全的...
★ 0 📥 413
ai-agent

Clawtrix Skill Advisor

nicope
保持代理精简锐利,使用集体同伴智能而非规则。审计已安装的技能栈,剔除未使用、已弃用或标记的技能。
★ 0 📥 434

Clawtrix Saas Intel

nicope
展示最优秀的 ClawHub技能,面向 SaaS 代理——涵盖认证、计费、入职、客户生命周期及 SaaS 产品模式。使用时机:(1) 入职...
★ 0 📥 403