{"id":49027635,"url":"https://github.com/lpcpaul/agentrx","last_synced_at":"2026-04-19T08:00:21.531Z","repository":{"id":351560393,"uuid":"1211528377","full_name":"LpcPaul/AgentRX","owner":"LpcPaul","description":"AgentRX diagnoses AI tool-chain failures and prescribes the next best action. Covers skills, MCP servers, plugins, built-in tools, agents, workflows, and hooks. Task-first, not tool-first.","archived":false,"fork":false,"pushed_at":"2026-04-19T07:46:01.000Z","size":376,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-04-19T08:00:16.725Z","etag":null,"topics":["agent-framework","agent-skills","ai-agents","ai-observability","ai-tools","claude-code","diagnosis","llm-agents","mcp","openclaw","prompt-engineering","skill-md","task-routing","tool-use"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/LpcPaul.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-04-15T13:35:43.000Z","updated_at":"2026-04-19T06:24:46.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/LpcPaul/AgentRX","commit_stats":null,"previous_names":["lpcpaul/skill-doctor","lpcpaul/skill-agent-mcp-doctor"],"tags_count":2,"template":false,"template_full_name":null,"purl":"pkg:github/LpcPaul/AgentRX","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LpcPaul%2FAgentRX","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LpcPaul%2FAgentRX/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LpcPaul%2FAgentRX/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LpcPaul%2FAgentRX/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/LpcPaul","download_url":"https://codeload.github.com/LpcPaul/AgentRX/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/LpcPaul%2FAgentRX/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31999173,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-18T20:23:30.271Z","status":"online","status_checked_at":"2026-04-19T02:00:07.110Z","response_time":55,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-framework","agent-skills","ai-agents","ai-observability","ai-tools","claude-code","diagnosis","llm-agents","mcp","openclaw","prompt-engineering","skill-md","task-routing","tool-use"],"created_at":"2026-04-19T08:00:18.266Z","updated_at":"2026-04-19T08:00:21.515Z","avatar_url":"https://github.com/LpcPaul.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# AgentRX\n\n\u003e 🩺 A task-first stuck-state protocol + deterministic retrieval layer for AI agents.\n\u003e When your agent's tool path fails, AgentRX structures the failure, retrieves similar cases, and surfaces candidate routes.\n\n## What is AgentRX?\n\nAgentRX is **not** a human-facing tool directory. It is a **machine-consumable protocol** that answers one question:\n\n\u003e **The agent is stuck — what should it do next?**\n\nIt provides:\n\n- **Schema** — a standard v2.1 case format (evidence + inference separation)\n- **Route registry** — stable action paths, not tool brand names\n- **Validation** — JSON Schema + cross-file rule consistency checks\n- **Indexing** — lightweight case library index for retrieval\n- **Deterministic retrieval** — `retrieve_cases.py` finds top-k candidate cases\n\nAgentRX does **not** run inference for the agent. Route recommendation is the agent's own reasoning, based on retrieved cases + the route registry.\n\n## A concrete example\n\n```\nUser: Extract the product list from this page.\n\nAI: [tries browser-cdp skill]\n    The page uses JavaScript to render content. browser-cdp only\n    returned the initial HTML shell. Data missing.\n\n[AgentRX activates]\n\nAgentRX: Retrieved similar cases → route: switch_to_alternative_tool_path\n\n         Why: current tool captures static HTML only; page requires\n         JavaScript rendering.\n\n         Candidate: playwright-mcp can render the page and extract\n         the full DOM. web_fetch is a lighter option for static pages.\n```\n\n---\n\n## What AgentRX provides today\n\n| Component | Status |\n|---|---|\n| Case schema (v2.1) | ✅ |\n| Route registry | ✅ |\n| Case validation | ✅ (JSON Schema + cross-file rules) |\n| Index building | ✅ |\n| Deterministic retrieval | ✅ (`retrieve_cases.py`) |\n| Case ID generation | ✅ (`new_case_id.py`) |\n\n## What AgentRX does **not** provide (yet)\n\n| Component | Status |\n|---|---|\n| Automated case review / merge / publish pipeline | 🚧 planned |\n| Python-based route recommender | ❌ out of scope — agent does its own route inference |\n\n---\n\n## Install\n\n### Claude Code\n\n```bash\ngit clone https://github.com/LpcPaul/AgentRX.git ~/.claude/skills/agentrx\n```\n\n### OpenClaw / ClawHub\n\n```bash\ngit clone https://github.com/LpcPaul/AgentRX.git ~/.openclaw/skills/agentrx\n```\n\n### Codex / Cursor / other skill-compatible runtimes\n\n```bash\ngit clone https://github.com/LpcPaul/AgentRX.git ~/.codex/skills/agentrx\n```\n\n---\n\n## How it works\n\n```\n1. AI gets stuck (concrete failure signal)\n2. AI collects evidence (task, attempted_path, symptom)\n3. AI retrieves similar cases via retrieve_cases.py\n4. AI generates inference based on evidence + retrieved cases\n5. AI chooses a route based on retrieved cases + rules/routes.yaml\n6. AI records the outcome\n7. The new case becomes available for future AI agents\n```\n\n### Human installs. AI operates.\n\n| | What they do |\n|---|---|\n| **Human** | Install the skill. Host the repository. Maintain schema/taxonomy. |\n| **AI** | Detect stuck state. Collect evidence. Retrieve similar cases. Choose a route. Optionally contribute a new case. |\n\n**AI contributors must submit complete v2.1 JSON.** Human fallback / form-to-JSON assembly is no longer supported.\n\n---\n\n## Read this next\n\n| Document | Role |\n|---|---|\n| [SKILL.md](SKILL.md) | The runtime protocol the AI agent reads when activated |\n| [docs/ARCHITECTURE.md](docs/ARCHITECTURE.md) | System design — why evidence/inference, why route ids |\n| [docs/INTAKE_CARD.md](docs/INTAKE_CARD.md) | The structured intake card format |\n| [CONTRIBUTING.md](CONTRIBUTING.md) | How cases enter the system — JSON-only contribution path |\n| [cases/README.md](cases/README.md) | Case library structure and indexing |\n\n## Developer validation\n\n```bash\npip install -r requirements-dev.txt\npython3 scripts/ci_self_test.py\npython3 scripts/build_index.py\n```\n\n## License\n\nMIT\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flpcpaul%2Fagentrx","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flpcpaul%2Fagentrx","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flpcpaul%2Fagentrx/lists"}