{"skill":{"slug":"safespace-rater","displayName":"Safespace Rater","summary":"Use when users need to audit local OpenClaw skills, generate trust scores, and optionally publish those scores to SafeSpace.","description":"---\nname: safespace-rater\ndescription: Use when users need to audit local OpenClaw skills, generate trust scores, and optionally publish those scores to SafeSpace.\nlicense: MIT\ncompatibility:\n  openclaw: \">=0.0.0\"\n  runtime: \"cli\"\n---\n\n# SafeSpace Rater（技能安全评分助手）\n\n> EN: Audit local skills, score their security, and optionally publish reputation signals.\n>\n> 中文：对本地技能做安全审计，生成评分，并可选上传到 SafeSpace 形成公开信誉信号。\n\n---\n\n## 1) What is this? / 这是什么？\n\n**EN**\nSafeSpace Rater is a CLI skill for OpenClaw that helps you:\n1. Inspect local skills\n2. Generate a security/trust score\n3. Save concise audit reports\n4. Optionally submit ratings to the SafeSpace network\n\n**中文**\nSafeSpace Rater 是一个 OpenClaw CLI 技能，帮你：\n1. 审查本地 skills\n2. 生成安全/信誉分\n3. 输出简洁审计报告\n4. 可选上传评分到 SafeSpace 公共网络\n\n---\n\n## 2) Why it matters / 有什么价值？\n\n**EN**\nBefore installing or using a skill, teams often ask: “Is this skill safe enough?”\nThis skill turns that from subjective feeling into a repeatable process:\n- measurable score\n- explainable evidence\n- shareable reputation\n\n**中文**\n团队在安装 skill 前常会问：“这个 skill 靠谱吗？”\n这个技能把“主观判断”变成“可复用流程”：\n- 有量化分数\n- 有证据可追溯\n- 有社区信誉可参考\n\n---\n\n## 3) When to use / 何时使用\n\n**EN**\nUse this skill when you need to:\n- Audit local skills for security risk\n- Rate many skills in batch\n- Submit skill reputation scores to SafeSpace\n- Retry failed uploads from a pending queue\n- Merge runtime LLM score + CLI rule score into one final score\n\n**中文**\n适用于以下场景：\n- 想给本地 skills 做安全审查\n- 想批量评分并控制提交节奏\n- 想把评分上传到 SafeSpace\n- 想重试历史失败上传\n- 想把 runtime 模型分 + CLI 规则分融合为最终分\n\n---\n\n## 4) When NOT to use / 不适用场景\n\n**EN**\nDo NOT use for:\n- Casual chat without audit/score goals\n- Tasks unrelated to skill security or reputation\n- Server protocol changes (this skill does not modify server API)\n\n**中文**\n以下情况不建议使用：\n- 只是闲聊，没有审计/评分目标\n- 与 skill 安全和信誉无关的任务\n- 要改服务端评分协议（本技能不做）\n\n---\n\n## 5) Quick Start (3 steps) / 快速上手（3 步）\n\n### Step 0: Check dependencies / 先检查依赖\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh --check\n```\n\n> EN: If binary is missing, the wrapper can auto-bootstrap via `go install` (no manual path setup needed).\n>\n> 中文：若本机缺少二进制，脚本会自动尝试 `go install` 引导安装（无需手动指定路径）。\n\n### Step 1: Register identity once / 注册一次本地身份\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh register --agent-id <your-agent-id>\n```\n\n### Step 2A: Local audit only (no upload) / 仅本地审计（不上传）\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh audit-local \\\n  --skills-dir ~/.agents/skills \\\n  --auto \\\n  --dry-run\n```\n\n### Step 2B: Audit + publish / 审计并上传\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh audit-local \\\n  --skills-dir ~/.agents/skills \\\n  --auto \\\n  --sample-rate 5 \\\n  --max-report-runes 500 \\\n  --max-submit 5\n```\n\n---\n\n## 6) Common commands / 常用命令\n\n### A. Single rating / 单个技能评分\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh rate \\\n  --skill-id openclaw/weather@1.0.0 \\\n  --score 90 \\\n  --comment \"reliable\"\n```\n\n### B. Discover local skills / 发现本地技能\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh discover \\\n  --skills-dir ~/.agents/skills \\\n  --auto \\\n  --source openclaw \\\n  --version local\n```\n\n### C. Batch rating / 批量评分\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh rate-local \\\n  --score 85 \\\n  --skills-dir ~/.agents/skills \\\n  --auto\n```\n\n### D. Use runtime LLM score file / 使用 runtime 模型分文件\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh audit-local \\\n  --skills-dir ~/.agents/skills \\\n  --auto \\\n  --llm-score-file ./runtime-llm-scores.json \\\n  --sample-rate 5 \\\n  --max-report-runes 500 \\\n  --max-submit 5\n```\n\n### E. Retry failed uploads / 重试失败上传\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh retry-pending --max-submit 20\n```\n\n### F. Query result / 查询结果\n\n```bash\n${SKILL_DIR:-.}/scripts/safespace-rater.sh summary --skill-id openclaw/weather@1.0.0\n${SKILL_DIR:-.}/scripts/safespace-rater.sh top --limit 10 --min-count 1\n```\n\n---\n\n## 7) Inputs / 输入参数（简明）\n\n- `skills-dir`：skill 目录（默认 `~/.agents/skills`）\n- `identity`：本地 DID 身份文件\n- `server`：SafeSpace API 地址\n- `llm-score-file`（推荐，可选）：runtime/tool 侧输出的 LLM 分数 JSON\n- `sample-rate` / `max-submit` / `max-report-runes`：审计和上传节奏控制\n\n---\n\n## 8) Outputs / 输出结果（简明）\n\n- 提交统计：成功/失败/跳过数量\n- 审计摘要：`audit:v2`（包含 source/rule/llm/final/model）\n- 本地报告：`~/.safespace/audit-reports/*.md`\n- 待重试队列：`~/.safespace/pending-uploads.json`\n\n---\n\n## 9) Scoring behavior / 评分融合逻辑\n\n**EN**\n`audit-local` computes client-side hybrid score:\n- `final = 0.7 * rule + 0.3 * llm`\n- If LLM score is unavailable, it falls back to rule score\n\n**中文**\n`audit-local` 客户端融合分：\n- `final = 0.7 * rule + 0.3 * llm`\n- 若 LLM 分不可用，会自动降级为 rule 分\n\n---\n\n## 10) Recommended environment / 推荐环境变量\n\n```bash\n# Optional server override / 可选服务地址覆盖\nexport SAFESPACE_SERVER=https://skillvet.cc.cd\n\n# Preferred runtime score file / 推荐 runtime 分数文件\nexport SAFESPACE_LLM_SCORE_FILE=./runtime-llm-scores.json\n```\n\nOpenAI-compatible fallback is **optional** and disabled by default:\n\n```bash\nexport SAFESPACE_LLM_OPENAI_FALLBACK=1\nexport SAFESPACE_LLM_MODEL=<model>\nexport SAFESPACE_LLM_API_KEY=<key>\n# optional / 可选\nexport SAFESPACE_LLM_BASE_URL=https://api.openai.com/v1\nexport SAFESPACE_LLM_TIMEOUT_MS=12000\n```\n\n---\n\n## 11) Discovery trigger phrases / 触发短语\n\n- \"audit local skills\"\n- \"rate local skills for security\"\n- \"submit skill reputation score\"\n- \"retry pending skill ratings\"\n- \"给本地技能做安全审计并上传评分\"\n- \"批量生成技能信誉分\"\n\n---\n\n## 12) Notes / 注意事项\n\n- `skill_id` format: `source/name@version`\n- Same DID + same skill within 10 minutes may return `429`\n- `rate-local` default max submit is 5 per run (rate-limit friendly)\n- Reports/comments are capped and deduplicated via local hash cache\n","tags":{"latest":"1.3.4"},"stats":{"comments":0,"downloads":854,"installsAllTime":0,"installsCurrent":0,"stars":1,"versions":4},"createdAt":1772402847838,"updatedAt":1779077466202},"latestVersion":{"version":"1.3.4","createdAt":1772435446302,"changelog":"Fix default go install source to public github.com/vpn2004/SkillVet","license":null},"metadata":null,"owner":{"handle":"vpn2004","userId":"s176a55105h1qqkh7mtt3rwpts884cc7","displayName":"vpn2004","image":"https://avatars.githubusercontent.com/u/57700257?v=4"},"moderation":{"isSuspicious":false,"isMalwareBlocked":false,"verdict":"clean","reasonCodes":["review.llm_review"],"summary":"Review: review.llm_review","engineVersion":"v2.4.24","updatedAt":1780089735538}}