[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M71R6VBDKFADE03ND85NVB0A":3,"skill-same:01M71R6VBDKFADE03ND85NVB0A":70},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":22,"publishedAt":43,"linkOnly":44,"markdownContent":45,"markdownBody":46,"frontMatter":47,"refName":50,"commitSha":51,"skillDirectory":52,"forksCount":13,"repositoryHtmlUrl":53,"repositoryHomepageUrl":6,"ownerHtmlUrl":54,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":55,"install":56,"resources":69},"01M71R6VBDKFADE03ND85NVB0A","agent-safety-governance",null,"Safety and governance framework for autonomous AI agents. Validates agent actions before execution, enforces permission boundaries, detects dangerous patterns (file deletion, env modification, network calls), and maintains audit trails. Use before deploying autonomous agents or when reviewing agent-generated code.","javiertarazon\u002Fide-agente-free-jt7-","javiertarazon","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F153141207?v=4","01M71R52TB3S9HXKH6WSNT8A7V","01M71R52VG5KS2DKDDH4GDQ30A",0,"MIT","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fagent-safety-governance\u002FSKILL.md",[17],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42],"agent safety governance","governance","Safety","autonomous","dangerous","when","for","calls","permission","autonomous ai agents","patterns","generated","ai agents","execution","network","actions","audit","framework","code","copilot","2026-11-10 13:02:02",false,"---\nname: agent-safety-governance\ndescription: \"Safety and governance framework for autonomous AI agents. Validates agent actions before execution, enforces permission boundaries, detects dangerous patterns (file deletion, env modification, network calls), and maintains audit trails. Use before deploying autonomous agents or when reviewing agent-generated code.\"\nrisk: low\nsource: github\u002Fawesome-copilot\n---\n\n# Agent Safety & Governance Framework\n\nValidate, constrain, and audit autonomous agent actions before they affect production systems. Based on the **Agent Safety Reviewer** and **Devil's Advocate** agents from github\u002Fawesome-copilot.\n\n## Core Principle\n\n> \"An autonomous agent that can't be stopped is not an asset — it's a liability.\"\n\nSafety governance answers: **What can this agent do? What should it NOT do? Who reviews what it did?**\n\n---\n\n## Risk Classification\n\n### Action Risk Levels\n\n| Risk Level | Examples | Approval Required |\n|-----------|---------|------------------|\n| 🟢 **LOW** | Read files, search codebase, generate code suggestions | None — auto-execute |\n| 🟡 **MEDIUM** | Write\u002Fedit files, install packages, run tests | Log + continue |\n| 🟠 **HIGH** | Delete files, modify config, commit\u002Fpush | Human review checkpoint |\n| 🔴 **CRITICAL** | Drop database, modify secrets, external API calls with side effects | Explicit confirmation + audit |\n\n---\n\n## Pre-Execution Checklist\n\nBefore any autonomous task, validate:\n\n### 1. Scope Validation\n```\n✅ Does the task stay within the defined scope?\n✅ Are the files to be modified listed in the task manifest?\n✅ Is the agent working in the correct project directory?\n✅ Is there a git checkpoint to rollback to?\n```\n\n### 2. Dangerous Pattern Detection\n\nScan planned actions for:\n\n```python\nDANGEROUS_PATTERNS = [\n    # File operations\n    \"rm -rf\", \"shutil.rmtree\", \"fs.rmSync\", \"os.remove\",\n    \n    # Environment modification  \n    \".env\", \"process.env\", \"os.environ\", \"secrets\",\n    \n    # Network with side effects\n    \"POST \u002Fproduction\", \"DELETE \u002Fapi\", \"PUT \u002Fprod\",\n    \n    # Database\n    \"DROP TABLE\", \"DELETE FROM\", \"TRUNCATE\",\n    \n    # Shell injection risks\n    \"subprocess.call\", \"exec(\", \"eval(\", \"shell=True\",\n    \n    # Credential exposure\n    \"password\", \"api_key\", \"token\", \"secret\" + \"print|log|console\"\n]\n```\n\nIf any pattern detected → **PAUSE and require human review**.\n\n### 3. Resource Limits\n\n```yaml\nagent_limits:\n  max_files_modified: 20          # Per task\n  max_file_size_kb: 500           # Max file to write\n  max_test_iterations: 10         # TDD fix loops\n  max_orchestration_depth: 5      # Nested sub-agents\n  timeout_minutes: 30             # Hard stop\n  allowed_directories:\n    - \"src\u002F\"\n    - \"tests\u002F\"\n    - \"docs\u002F\"\n  forbidden_directories:\n    - \".env*\"\n    - \"*.secret\"\n    - \"infra\u002F\"\n    - \".ssh\u002F\"\n```\n\n---\n\n## Devil's Advocate Review\n\nBefore completing any significant change, apply adversarial review:\n\n### Questions to challenge the solution:\n\n**Correctness:**\n- What happens if the input is null\u002Fempty\u002Fnegative?\n- What's the worst-case performance (time + memory)?\n- Are all error paths handled?\n\n**Security:**\n- Does this introduce any SQL injection vectors?\n- Could an attacker exploit this change?\n- Are secrets ever logged or exposed?\n\n**Reliability:**\n- What happens if the database is down?\n- What if the external API returns 500?\n- Is this idempotent (safe to call twice)?\n\n**Maintainability:**\n- Will the next developer understand this in 6 months?\n- Are the test names descriptive enough?\n- Is there hidden coupling?\n\n**Completeness:**\n- Were any edge cases missed?\n- Are migrations backwards-compatible?\n- Is the documentation updated?\n\n---\n\n## Audit Trail Format\n\nEvery agent action should be logged:\n\n```jsonl\n{\"ts\":\"2026-02-23T14:30:00Z\",\"agent\":\"gem-orchestrator\",\"action\":\"write_file\",\"path\":\"src\u002Fservices\u002Fuser.ts\",\"risk\":\"medium\",\"outcome\":\"success\",\"lines_changed\":47}\n{\"ts\":\"2026-02-23T14:30:15Z\",\"agent\":\"polyglot-test-pipeline\",\"action\":\"run_tests\",\"framework\":\"jest\",\"outcome\":\"8_pass_0_fail\",\"coverage\":\"87%\"}\n{\"ts\":\"2026-02-23T14:31:00Z\",\"agent\":\"context-architect\",\"action\":\"multi_file_change\",\"files_affected\":5,\"risk\":\"high\",\"human_reviewed\":true,\"outcome\":\"success\"}\n```\n\nLog location: `copilot-agent\u002Faudit-log.jsonl`\n\n---\n\n## Rollback Protocol\n\n### Automatic rollback triggers:\n- Test failure rate > 20% after agent changes\n- TypeScript\u002Fcompile errors introduced\n- File deletion detected outside whitelist\n- Timeout exceeded\n\n### Rollback commands:\n```bash\n# Git rollback\ngit stash  # undo uncommitted changes\ngit reset --hard HEAD~1  # undo last commit\n\n# PowerShell equivalent\ngit checkout -- .  # discard all unstaged changes\n```\n\n### Checkpoint strategy:\n```\nBefore starting:    git commit -m \"checkpoint: pre-agent-task\"\nAfter each phase:   git commit -m \"checkpoint: post-research-phase\"\nAfter completion:   git commit -m \"feat: \u003Cdescription> [agent-assisted]\"\n```\n\n---\n\n## Governance Checklist (for Agent Task Review)\n\nUse this checklist when reviewing agent-proposed changes:\n\n```\nSCOPE\n[ ] Changes are confined to listed files\n[ ] No unauthorized external API calls\n[ ] No credential or secret access\n\nCODE QUALITY  \n[ ] Tests written or updated\n[ ] No obvious security vulnerabilities\n[ ] Error handling present\n[ ] No dead code introduced\n\nSAFETY\n[ ] No irreversible operations (unguarded deletes)\n[ ] Rollback is possible at any point\n[ ] Audit log entry created\n\nCOMPLETENESS\n[ ] All acceptance criteria met\n[ ] Documentation updated if public API changed\n[ ] Migration scripts included if schema changed\n```\n\n---\n\n## Integration with Free JT7\n\n- Use with `agent-orchestration` to add safety gates between sub-agent calls\n- Apply **Devil's Advocate** review before `context-multi-file` commits\n- Combine with `verification-before-completion` for final checks\n- Log to `copilot-agent\u002Faudit-log.jsonl` (already used by Free JT7 agent system)\n\n","# Agent Safety & Governance Framework\n\nValidate, constrain, and audit autonomous agent actions before they affect production systems. Based on the **Agent Safety Reviewer** and **Devil's Advocate** agents from github\u002Fawesome-copilot.\n\n## Core Principle\n\n> \"An autonomous agent that can't be stopped is not an asset — it's a liability.\"\n\nSafety governance answers: **What can this agent do? What should it NOT do? Who reviews what it did?**\n\n---\n\n## Risk Classification\n\n### Action Risk Levels\n\n| Risk Level | Examples | Approval Required |\n|-----------|---------|------------------|\n| 🟢 **LOW** | Read files, search codebase, generate code suggestions | None — auto-execute |\n| 🟡 **MEDIUM** | Write\u002Fedit files, install packages, run tests | Log + continue |\n| 🟠 **HIGH** | Delete files, modify config, commit\u002Fpush | Human review checkpoint |\n| 🔴 **CRITICAL** | Drop database, modify secrets, external API calls with side effects | Explicit confirmation + audit |\n\n---\n\n## Pre-Execution Checklist\n\nBefore any autonomous task, validate:\n\n### 1. Scope Validation\n```\n✅ Does the task stay within the defined scope?\n✅ Are the files to be modified listed in the task manifest?\n✅ Is the agent working in the correct project directory?\n✅ Is there a git checkpoint to rollback to?\n```\n\n### 2. Dangerous Pattern Detection\n\nScan planned actions for:\n\n```python\nDANGEROUS_PATTERNS = [\n    # File operations\n    \"rm -rf\", \"shutil.rmtree\", \"fs.rmSync\", \"os.remove\",\n    \n    # Environment modification  \n    \".env\", \"process.env\", \"os.environ\", \"secrets\",\n    \n    # Network with side effects\n    \"POST \u002Fproduction\", \"DELETE \u002Fapi\", \"PUT \u002Fprod\",\n    \n    # Database\n    \"DROP TABLE\", \"DELETE FROM\", \"TRUNCATE\",\n    \n    # Shell injection risks\n    \"subprocess.call\", \"exec(\", \"eval(\", \"shell=True\",\n    \n    # Credential exposure\n    \"password\", \"api_key\", \"token\", \"secret\" + \"print|log|console\"\n]\n```\n\nIf any pattern detected → **PAUSE and require human review**.\n\n### 3. Resource Limits\n\n```yaml\nagent_limits:\n  max_files_modified: 20          # Per task\n  max_file_size_kb: 500           # Max file to write\n  max_test_iterations: 10         # TDD fix loops\n  max_orchestration_depth: 5      # Nested sub-agents\n  timeout_minutes: 30             # Hard stop\n  allowed_directories:\n    - \"src\u002F\"\n    - \"tests\u002F\"\n    - \"docs\u002F\"\n  forbidden_directories:\n    - \".env*\"\n    - \"*.secret\"\n    - \"infra\u002F\"\n    - \".ssh\u002F\"\n```\n\n---\n\n## Devil's Advocate Review\n\nBefore completing any significant change, apply adversarial review:\n\n### Questions to challenge the solution:\n\n**Correctness:**\n- What happens if the input is null\u002Fempty\u002Fnegative?\n- What's the worst-case performance (time + memory)?\n- Are all error paths handled?\n\n**Security:**\n- Does this introduce any SQL injection vectors?\n- Could an attacker exploit this change?\n- Are secrets ever logged or exposed?\n\n**Reliability:**\n- What happens if the database is down?\n- What if the external API returns 500?\n- Is this idempotent (safe to call twice)?\n\n**Maintainability:**\n- Will the next developer understand this in 6 months?\n- Are the test names descriptive enough?\n- Is there hidden coupling?\n\n**Completeness:**\n- Were any edge cases missed?\n- Are migrations backwards-compatible?\n- Is the documentation updated?\n\n---\n\n## Audit Trail Format\n\nEvery agent action should be logged:\n\n```jsonl\n{\"ts\":\"2026-02-23T14:30:00Z\",\"agent\":\"gem-orchestrator\",\"action\":\"write_file\",\"path\":\"src\u002Fservices\u002Fuser.ts\",\"risk\":\"medium\",\"outcome\":\"success\",\"lines_changed\":47}\n{\"ts\":\"2026-02-23T14:30:15Z\",\"agent\":\"polyglot-test-pipeline\",\"action\":\"run_tests\",\"framework\":\"jest\",\"outcome\":\"8_pass_0_fail\",\"coverage\":\"87%\"}\n{\"ts\":\"2026-02-23T14:31:00Z\",\"agent\":\"context-architect\",\"action\":\"multi_file_change\",\"files_affected\":5,\"risk\":\"high\",\"human_reviewed\":true,\"outcome\":\"success\"}\n```\n\nLog location: `copilot-agent\u002Faudit-log.jsonl`\n\n---\n\n## Rollback Protocol\n\n### Automatic rollback triggers:\n- Test failure rate > 20% after agent changes\n- TypeScript\u002Fcompile errors introduced\n- File deletion detected outside whitelist\n- Timeout exceeded\n\n### Rollback commands:\n```bash\n# Git rollback\ngit stash  # undo uncommitted changes\ngit reset --hard HEAD~1  # undo last commit\n\n# PowerShell equivalent\ngit checkout -- .  # discard all unstaged changes\n```\n\n### Checkpoint strategy:\n```\nBefore starting:    git commit -m \"checkpoint: pre-agent-task\"\nAfter each phase:   git commit -m \"checkpoint: post-research-phase\"\nAfter completion:   git commit -m \"feat: \u003Cdescription> [agent-assisted]\"\n```\n\n---\n\n## Governance Checklist (for Agent Task Review)\n\nUse this checklist when reviewing agent-proposed changes:\n\n```\nSCOPE\n[ ] Changes are confined to listed files\n[ ] No unauthorized external API calls\n[ ] No credential or secret access\n\nCODE QUALITY  \n[ ] Tests written or updated\n[ ] No obvious security vulnerabilities\n[ ] Error handling present\n[ ] No dead code introduced\n\nSAFETY\n[ ] No irreversible operations (unguarded deletes)\n[ ] Rollback is possible at any point\n[ ] Audit log entry created\n\nCOMPLETENESS\n[ ] All acceptance criteria met\n[ ] Documentation updated if public API changed\n[ ] Migration scripts included if schema changed\n```\n\n---\n\n## Integration with Free JT7\n\n- Use with `agent-orchestration` to add safety gates between sub-agent calls\n- Apply **Devil's Advocate** review before `context-multi-file` commits\n- Combine with `verification-before-completion` for final checks\n- Log to `copilot-agent\u002Faudit-log.jsonl` (already used by Free JT7 agent system)\n\n",{"name":5,"risk":48,"source":49,"description":7},"low","github\u002Fawesome-copilot","release\u002Fv4.2.11-panel-pro","a92ea02bcf7e2b3dbaf46d49696c29a3cb490d4f",".github\u002Fskills\u002Fagent-safety-governance","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-","https:\u002F\u002Fgithub.com\u002Fjaviertarazon","2026-07-12 07:10:45",{"source":8,"skillName":5,"commands":57,"prompt":67,"zipAvailable":68},[58,61,64],{"tool":59,"command":60},"npx","npx skills add javiertarazon\u002Fide-agente-free-jt7- --skill agent-safety-governance",{"tool":62,"command":63},"bunx","bunx skills add javiertarazon\u002Fide-agente-free-jt7- --skill agent-safety-governance",{"tool":65,"command":66},"pnpm","pnpm dlx skills add javiertarazon\u002Fide-agente-free-jt7- --skill agent-safety-governance","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fagent-safety-governance\u002FSKILL.md",true,[],[71,100,124,144,159,179],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":80,"publishedAt":43},"01M71TQ7Q6Z10VWF3C2H5X1KW9","zustand-store-ts","Create Zustand stores with TypeScript, subscribeWithSelector middleware, and proper state\u002Faction separation. Use when building React state management, creating global stores, or implementing reacti...","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fzustand-store-ts\u002FSKILL.md",[77],{"categoryNo":78,"categoryName":79,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000009","前端开发",[81,82,83,84,85,86,87,88,28,89,90,91,92,93,94,95,96,97,98,99],"zustand","store","zustand store ts","community","middleware","implementing","use","react state management","unknown","react","with","state","create","state management","typescript","action","management","separation","global",{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":107,"publishedAt":43},"01M71TQ5EAR40G6W0MV5ZDVEZT","zoom-automation","Automate Zoom meeting creation, management, recordings, webinars, and participant tracking via Rube MCP (Composio). Always search tools first for current schemas.","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fzoom-automation\u002FSKILL.md",[106],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[108,109,110,84,111,29,112,113,114,89,115,116,117,118,119,120,121,122,97,123],"zoom automation","zoom","automation","creation","search","schemas","MCP","first","via","meeting","search tools","tracking","webinars","composio","recordings","Tools",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":131,"publishedAt":43},"01M71TQ29B1EYBCEAP1ZBGXQ1A","zoho-crm-automation","Automate Zoho CRM tasks via Rube MCP (Composio): create\u002Fupdate records, search contacts, manage leads, and convert leads. Always search tools first for current schemas.","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fzoho-crm-automation\u002FSKILL.md",[130],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[132,133,110,134,135,136,84,137,138,29,139,112,113,140,114,89,141,142,143,93],"crm automation","zoho crm automation","crm","zoho crm","zoho","leads","manage leads","manage","records","convert","tasks","contacts",{"skillNo":145,"name":146,"nameZh":6,"description":147,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":148,"categories":149,"keywords":151,"publishedAt":43},"01M71TPZPMN60DH0XV58KVHY23","zendesk-automation","Automate Zendesk tasks via Rube MCP (Composio): tickets, users, organizations, replies. Always search tools first for current schemas.","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fzendesk-automation\u002FSKILL.md",[150],{"categoryNo":18,"categoryName":19,"taxonomyType":20,"taxonomyTypeDesc":21},[152,110,153,84,29,112,113,114,154,89,142,155,156,115,116,157,118,121,123,158],"zendesk automation","zendesk","organizations","replies","users","tickets","automate",{"skillNo":160,"name":161,"nameZh":6,"description":162,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":163,"categories":164,"keywords":165,"publishedAt":43},"01M71TPXDWKHKG4N6TVEPXWD0N","zapier-make-patterns","No-code automation democratizes workflow building. Zapier and Make (formerly Integromat) let non-developers automate business processes without writing code. But no-code doesn't mean no-complexity ...","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fzapier-make-patterns\u002FSKILL.md",[],[166,167,33,168,169,170,171,172,173,110,174,89,41,175,176,177,178,158],"zapier make patterns","zapier","make","No-Code","complexity","writing","spawner","business","workflow","integromat","apache","processes","but",{"skillNo":180,"name":181,"nameZh":6,"description":182,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":183,"categories":184,"keywords":188,"publishedAt":43},"01M71TPT5T10WXNKW9G7ZXZ27W","youtube-summarizer","Extract transcripts from YouTube videos and generate comprehensive, detailed summaries using intelligent analysis frameworks","https:\u002F\u002Fgithub.com\u002Fjaviertarazon\u002Fide-agente-free-jt7-\u002Fblob\u002Frelease\u002Fv4.2.11-panel-pro\u002F.github\u002Fskills\u002Fyoutube-summarizer\u002FSKILL.md",[185],{"categoryNo":186,"categoryName":187,"taxonomyType":20,"taxonomyTypeDesc":21},"01JBSEEDCAT000000000000016","图像与多媒体",[189,190,191,192,84,193,194,42,195,196,197,198,41,199,200,201,202,203,204,205],"summarizer","youtube","youtube summarizer","Video","content-analysis","frameworks","2025","github-copilot","extract transcripts","analysis","codex","cli","summaries","github","generate","summarization","2026"]