菜单
Run the 19-intent satisfaction scorer across all agents, write report to Chartroom, and trigger self-healing actions on threshold breaches
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Supernor/openclaw-skills --skill agent-satisfaction命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
How Failsafe judges the weekly backup beat from PUSHED receipts — interpretation table, thresholds, and the coverage-gap question technique. Use every weekly beat.
Route ideas to The Workshop (Scribe on Telegram). Creates Intake topics for structured intent capture.
Audit existing skills against create-skill best practices. Finds gaps, scores quality, and produces upgrade recommendations.
基于 SOC 职业分类
正在显示 SKILL.md
| name | agent-satisfaction |
| description | Run the 19-intent satisfaction scorer across all agents, write report to Chartroom, and trigger self-healing actions on threshold breaches |
| version | 1.0.0 |
| author | captain |
| tags | ["satisfaction","scoring","intents","health","self-healing"] |
| intent | Observable [I13], Reliable [I05] |
Score all 16 agents across 19 intents. Write results to Chartroom as report-agent-satisfaction. Trigger self-healing actions on threshold breaches.
Use MCP tools (works from inside container — no host paths needed):
satisfaction_scores MCP tool (no args for all agents)satisfaction_scores MCP tool with agent: "<agent-id>"satisfaction_summary MCP toolFallback (host only):
python3 /root/.openclaw/scripts/agent-satisfaction-score.py --json
satisfaction_scores MCP toolchart update report-agent-satisfaction "<summary>"| Condition | Action |
|---|---|
| Any agent context > 85% | Reset session via openclaw sessions --agent <id> --reset |
| Any agent context > 70% recurring (3+ snapshots) | Flag for workload evaluation |
| I03 Competent < 5 | Agent needs skills — flag for Dev |
| I02 Understood < 5 | SOUL.md needs rewrite — flag for Reactor |
| I08 Resilient < 5 | Missing fallback model — fix auth-profiles |
| I05 Reliable < 4 | Chronic errors — investigate model/config |
| Fleet avg < 6.0 | System-wide alert to Reactor |
Fleet avg: X.X/10 (16 agents). [date].
Top: [agent]=X.X. Bottom: [agent]=X.X.
Alerts: [threshold breaches].
Actions taken: [self-healing actions].
Captain checks report-agent-satisfaction date on first turn each day.
If older than 7 days (or missing), run this skill.