{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"github-zhayujie--cowagent","slug":"zhayujie--cowagent","name":"CowAgent","type":"agent","url":"https://cowagent.ai","page_url":"https://unfragile.ai/zhayujie--cowagent","categories":["chatbots-assistants"],"tags":["ai","ai-agent","chatgpt-on-wechat","claude","deepseek","dingtalk","feishu-bot","gemini","kimi","linkai","llm","mcp","multi-agent","openai","openclaw","python3","qwen","skills","wechat","weixin"],"pricing":{"model":"open_source","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"github-zhayujie--cowagent__cap_0","uri":"capability://tool.use.integration.multi.channel.message.routing.and.transformation","name":"multi-channel message routing and transformation","description":"CowAgent implements a ChannelFactory and ChannelManager pattern that abstracts communication platforms (WeChat, Feishu, DingTalk, WeCom, QQ, web console) into a unified message pipeline. Messages from heterogeneous sources are normalized into internal Context objects, routed through a Bridge component, and dispatched to appropriate Bot/Agent handlers running in separate daemon threads. This decouples platform-specific protocol handling from core reasoning logic, enabling concurrent multi-channel operation without cross-channel interference.","intents":["Deploy a single AI agent across multiple communication platforms simultaneously","Handle incoming messages from WeChat, enterprise platforms, and web interfaces without duplicating business logic","Scale message processing by running channel listeners in isolated threads"],"best_for":["Teams building enterprise digital employees across WeChat, Feishu, DingTalk ecosystems","Developers deploying personal AI assistants to multiple platforms from one codebase"],"limitations":["Channel-specific features (e.g., WeChat group mentions, Feishu card formatting) require custom plugin handlers","Message ordering guarantees only within a single channel; cross-channel consistency requires application-level coordination","Daemon thread model adds ~50-200ms latency per channel due to thread context switching"],"requires":["Python 3.7+","Platform-specific credentials (WeChat token, Feishu app ID, DingTalk webhook, etc.)","Network connectivity to target platforms"],"input_types":["text","voice","image","file"],"output_types":["text","voice","image","structured reply objects"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_1","uri":"capability://planning.reasoning.autonomous.task.planning.and.multi.step.execution","name":"autonomous task planning and multi-step execution","description":"CowAgent implements an Agent Execution Engine that decomposes user objectives into executable steps via chain-of-thought reasoning. The engine maintains a Prompt Builder that constructs context-aware prompts including available tools, memory, and workspace state. It iteratively invokes the LLM, parses tool-calling responses, executes tools (browser automation, terminal commands, skill invocations), and feeds results back into the reasoning loop until the goal is achieved. This creates a closed-loop planning system where the agent can autonomously decide which tools to invoke and when to stop.","intents":["Ask the agent to complete complex multi-step tasks without specifying each step","Enable the agent to autonomously decide which tools and skills to use based on task context","Allow the agent to recover from tool failures by replanning and retrying"],"best_for":["Solo developers building autonomous LLM agents for personal productivity","Non-technical users who want to delegate complex workflows to an AI assistant"],"limitations":["Planning quality depends on LLM reasoning capability; weaker models (e.g., GPT-3.5) may fail on complex multi-step tasks","No built-in cost control; unbounded tool invocations can lead to high API bills or long execution times","Agent execution is synchronous; long-running tasks block the message handler thread","No formal verification of plan correctness; agent may execute harmful commands if not properly constrained"],"requires":["API key for at least one LLM provider (OpenAI, Claude, Gemini, DeepSeek, Qwen, etc.)","Tools/skills registered in the agent's tool registry","Sufficient LLM context window to hold task description + available tools + execution history"],"input_types":["text (task description)"],"output_types":["text (final result)","structured execution trace"],"categories":["planning-reasoning","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_10","uri":"capability://automation.workflow.docker.containerization.and.cloud.deployment","name":"docker containerization and cloud deployment","description":"CowAgent provides Docker support through docker-compose configuration and container-ready deployment scripts. The system can be deployed as a containerized service, enabling easy scaling, version management, and cloud deployment. The Docker setup includes configuration for environment variables, volume mounts for persistence, and networking for multi-container deployments. CowAgent also integrates with LinkAI cloud platform for managed deployment and monitoring, providing an alternative to self-hosted deployment.","intents":["Deploy CowAgent as a containerized service in Docker or Kubernetes","Scale agent deployments across multiple containers with load balancing","Use managed cloud deployment via LinkAI platform without managing infrastructure"],"best_for":["DevOps teams deploying agents to production environments","Teams using Kubernetes or Docker Swarm for container orchestration","Non-technical teams that prefer managed cloud deployment over self-hosting"],"limitations":["Docker deployment requires Docker and docker-compose installation; adds operational complexity","Containerized agents cannot access host system resources (e.g., terminal commands) without explicit volume mounts or privileged mode","Cloud deployment via LinkAI requires vendor lock-in; migrating away requires re-deployment","Persistent storage (SQLite, skills directory) requires volume mounts; distributed deployments need shared storage"],"requires":["Docker and docker-compose installed","Docker image built from Dockerfile (or pulled from registry)","Environment variables configured for LLM providers and channels","Persistent volumes for SQLite and skills directory (if using stateful deployment)"],"input_types":["docker-compose.yml configuration","environment variables"],"output_types":["running container","container logs","health status"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_11","uri":"capability://image.visual.multi.modal.message.handling.with.image.and.file.processing","name":"multi-modal message handling with image and file processing","description":"CowAgent implements multi-modal message handling that processes text, voice, images, and files from various channels. The system includes image analysis capabilities (via vision-enabled LLMs like GPT-4V or Claude Vision) and file processing (e.g., PDF extraction, document parsing). Messages are normalized into a unified format regardless of source channel, and multi-modal content is passed to the LLM with appropriate encoding. This enables the agent to understand and respond to images, documents, and other non-text content.","intents":["Enable the agent to analyze images and screenshots provided by users","Process documents (PDFs, Word files) and extract information","Handle multi-modal conversations mixing text, images, and files"],"best_for":["Teams building agents that need to understand visual content (e.g., document processing, image analysis)","Developers creating assistants for knowledge workers who frequently share documents and screenshots"],"limitations":["Image analysis requires vision-capable LLM (e.g., GPT-4V, Claude Vision); not all providers support this","File processing requires format-specific parsers (e.g., PDF libraries); not all formats are supported","Large files (e.g., high-resolution images, multi-page PDFs) can exceed LLM context limits","Image analysis adds latency (~2-5 seconds per image) and cost (higher token usage)"],"requires":["Vision-capable LLM provider (OpenAI GPT-4V, Claude 3 Vision, Gemini Pro Vision, etc.)","File processing libraries (PyPDF2, python-docx, etc.) for supported formats","Sufficient LLM context window to accommodate image embeddings and file content"],"input_types":["text","image (JPEG, PNG, GIF, WebP)","file (PDF, DOCX, TXT, etc.)","voice"],"output_types":["text","structured data extracted from images/files"],"categories":["image-visual","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_12","uri":"capability://automation.workflow.configuration.management.with.template.based.setup","name":"configuration management with template-based setup","description":"CowAgent uses a configuration-driven approach with a config-template.json file that defines all agent settings (LLM provider, channels, plugins, memory, voice providers, etc.). The system loads configuration at startup and validates it against a schema. Users can customize behavior by editing the configuration file without modifying code. The configuration system supports environment variable substitution for sensitive values (API keys) and allows multiple configuration profiles for different deployment scenarios (development, staging, production).","intents":["Configure agent behavior (LLM provider, channels, plugins) without code changes","Manage sensitive credentials (API keys) via environment variables","Support multiple deployment profiles (dev, staging, prod) with different configurations"],"best_for":["Teams deploying agents to multiple environments with different configurations","Non-technical users who want to customize agent behavior via configuration files"],"limitations":["Configuration validation is basic; invalid configurations may not be caught until runtime","No built-in configuration versioning; changes to config-template.json can break existing deployments","Complex configurations become hard to manage; no UI for configuration management","Environment variable substitution is simple string replacement; no support for complex variable transformations"],"requires":["JSON configuration file (config.json or config-template.json)","Environment variables set for sensitive values (API keys, tokens)","Understanding of configuration schema and available options"],"input_types":["JSON configuration file","environment variables"],"output_types":["parsed configuration object","validation errors"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_2","uri":"capability://tool.use.integration.skill.hub.with.git.based.and.natural.language.installation","name":"skill hub with git-based and natural-language installation","description":"CowAgent provides a Skill Hub system that allows users to extend agent capabilities by installing new skills via Git repositories or natural-language dialogue. Skills are Python modules that register themselves as callable tools in the agent's tool registry. The system supports both explicit Git cloning (for developers) and conversational skill discovery (for non-technical users). Installed skills are persisted in a local skills directory and automatically loaded on agent startup, enabling rapid capability expansion without code modification.","intents":["Install new agent capabilities (e.g., web scraping, data analysis) without modifying core code","Share reusable skills across teams via Git repositories","Discover and install skills through natural conversation with the agent"],"best_for":["Teams building extensible AI assistants with community-contributed skills","Non-technical users who want to add capabilities through dialogue"],"limitations":["No sandboxing; installed skills run with full agent privileges and can access OS resources","No dependency resolution; skill conflicts or missing dependencies must be resolved manually","Natural-language skill discovery relies on LLM understanding of skill descriptions; may fail for poorly documented skills","Skills are loaded at startup; dynamic skill installation requires agent restart"],"requires":["Git installed and configured (for Git-based installation)","Write access to local skills directory","Python 3.7+ with pip for skill dependency installation"],"input_types":["text (skill name or Git URL)","natural language (skill discovery query)"],"output_types":["skill metadata","installation status","tool registry updates"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_3","uri":"capability://memory.knowledge.long.term.memory.with.temporal.decay.and.vector.retrieval","name":"long-term memory with temporal decay and vector retrieval","description":"CowAgent implements a dual-layer memory system that persists conversation history into local SQLite databases and vector stores. The system supports temporal decay scoring (older memories have lower relevance) and keyword-based retrieval alongside semantic vector search. Memory is organized by conversation context and can be queried to augment the agent's prompt with relevant historical information. This enables the agent to learn from past interactions and maintain continuity across sessions without relying on external knowledge bases.","intents":["Enable the agent to remember past conversations and user preferences across sessions","Retrieve relevant historical context to inform current task planning","Track long-term patterns in user behavior and adapt agent responses accordingly"],"best_for":["Personal AI assistants that need to maintain user context over weeks/months","Enterprise digital employees that must remember customer interaction history"],"limitations":["SQLite storage is not distributed; memory is local to the agent instance and not shared across deployments","Vector search requires embedding generation; adds ~100-500ms latency per memory query depending on vector store size","Temporal decay scoring is configurable but not adaptive; cannot learn optimal decay rates from usage patterns","No built-in privacy controls; all conversation history is stored locally in plaintext (or with basic encryption)"],"requires":["SQLite3 (included in Python standard library)","Vector embedding model (local or API-based) for semantic search","Sufficient disk space for conversation history (typically <1GB per 100k messages)"],"input_types":["text (conversation history)","embeddings (for vector search)"],"output_types":["retrieved memory snippets","augmented prompts with historical context"],"categories":["memory-knowledge","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_4","uri":"capability://tool.use.integration.multi.model.provider.abstraction.with.unified.interface","name":"multi-model provider abstraction with unified interface","description":"CowAgent abstracts LLM provider differences (OpenAI, Azure, Claude, Gemini, DeepSeek, Qwen, GLM, Kimi, LinkAI) behind a unified interface. The system implements provider-specific adapters that handle authentication, request formatting, response parsing, and error handling. Users can switch between providers via configuration without code changes. The abstraction layer also handles provider-specific features like function calling, vision capabilities, and streaming responses, normalizing them into a consistent API.","intents":["Switch between LLM providers (OpenAI, Claude, Gemini) without rewriting agent code","Use Chinese LLM providers (Qwen, DeepSeek, GLM) alongside Western providers","Leverage provider-specific features (e.g., Claude's extended context window) transparently"],"best_for":["Teams evaluating multiple LLM providers and wanting to avoid vendor lock-in","Developers building agents for Chinese markets who need native support for Qwen, DeepSeek, etc.","Cost-conscious teams that want to switch providers based on pricing or availability"],"limitations":["Provider-specific features (e.g., Claude's tool_choice parameter) may not be fully exposed through the abstraction","Response latency varies significantly by provider; no built-in load balancing or failover","Prompt engineering may need adjustment per provider due to different instruction-following capabilities","Token counting is provider-specific; abstraction layer cannot accurately predict costs across providers"],"requires":["API key for at least one supported LLM provider","Network connectivity to provider endpoints","Configuration file specifying provider and model selection"],"input_types":["text (prompts)","structured tool definitions"],"output_types":["text (completions)","structured tool calls","streaming tokens"],"categories":["tool-use-integration","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_5","uri":"capability://tool.use.integration.voice.processing.with.multi.provider.speech.to.text.and.text.to.speech","name":"voice processing with multi-provider speech-to-text and text-to-speech","description":"CowAgent integrates voice processing capabilities through a Voice Provider abstraction layer that supports multiple speech-to-text (STT) and text-to-speech (TTS) providers. The system can receive voice messages from channels (e.g., WeChat voice messages), transcribe them using configured STT providers, process the transcription through the agent, and synthesize responses back to voice using TTS providers. This enables fully voice-driven interaction with the agent across supported channels.","intents":["Interact with the agent using voice messages instead of text","Receive voice responses from the agent in natural-sounding speech","Build voice-first AI assistants for accessibility and hands-free operation"],"best_for":["Personal AI assistants for mobile and voice-first interfaces","Accessibility-focused applications for users with visual impairments"],"limitations":["STT accuracy depends on audio quality and background noise; no built-in noise cancellation","TTS latency adds 1-3 seconds per response; not suitable for real-time conversational interaction","Voice provider APIs have usage limits and per-minute costs; can become expensive at scale","Language support varies by provider; not all providers support all languages or accents"],"requires":["API key for at least one STT provider (e.g., OpenAI Whisper, Azure Speech Services)","API key for at least one TTS provider (e.g., OpenAI TTS, Azure Speech Services)","Audio codec support in the channel implementation (e.g., WeChat voice message format)"],"input_types":["audio (voice messages in channel-specific formats)"],"output_types":["text (transcriptions)","audio (synthesized speech)"],"categories":["tool-use-integration","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_6","uri":"capability://tool.use.integration.plugin.system.with.administrative.and.behavioral.plugins","name":"plugin system with administrative and behavioral plugins","description":"CowAgent implements a plugin architecture that allows extending agent behavior through administrative plugins (e.g., command handling, user management) and behavioral plugins (e.g., content filtering, response formatting). Plugins hook into the message pipeline at defined extension points (pre-processing, post-processing, tool invocation) and can modify context, intercept messages, or inject custom logic. The system loads plugins from a plugins directory at startup and maintains a plugin registry for runtime introspection.","intents":["Add custom command handlers (e.g., /help, /status) without modifying core code","Implement content filtering or response formatting policies","Extend agent behavior with domain-specific logic (e.g., customer service workflows)"],"best_for":["Teams building customized AI assistants with organization-specific behaviors","Developers who need to inject custom logic at specific points in the message pipeline"],"limitations":["Plugin API is not versioned; breaking changes in core code can break existing plugins","No plugin isolation; a misbehaving plugin can crash the entire agent","Plugin execution is synchronous; slow plugins block the message handler","No built-in plugin dependency management; plugins must explicitly declare their dependencies"],"requires":["Python 3.7+ with ability to import custom modules","Understanding of CowAgent's plugin API and extension points","Write access to the plugins directory"],"input_types":["Context objects (message, user, channel metadata)"],"output_types":["modified Context objects","side effects (logging, external API calls)"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_7","uri":"capability://tool.use.integration.web.console.channel.with.browser.based.interface","name":"web console channel with browser-based interface","description":"CowAgent includes a built-in Web Console channel that provides a browser-based interface for interacting with the agent. The console is implemented as a lightweight HTTP server that serves a web UI and handles WebSocket or HTTP polling for message exchange. This enables users to interact with the agent through a web browser without requiring platform-specific clients (e.g., WeChat app). The console supports the same multi-modal capabilities as other channels (text, voice, images).","intents":["Provide a web-based interface for testing and interacting with the agent","Enable users without WeChat or other platform accounts to access the agent","Build a custom web UI for the agent without implementing a separate frontend"],"best_for":["Developers prototyping and testing agents during development","Teams deploying agents to users who prefer web interfaces over platform-specific apps"],"limitations":["Web console is not production-hardened; no built-in authentication, rate limiting, or DDoS protection","Browser compatibility varies; older browsers may not support WebSocket or modern JavaScript features","No built-in persistence of web console sessions; messages are lost on page refresh unless explicitly saved","Scaling web console to many concurrent users requires load balancing and session management"],"requires":["Python 3.7+ with Flask or similar web framework","Network port available for HTTP server (default 5000 or configurable)","Modern web browser with JavaScript support"],"input_types":["text","voice","image","file (via file upload)"],"output_types":["text","voice","image","structured UI elements"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_8","uri":"capability://planning.reasoning.context.aware.prompt.building.with.workspace.and.tool.registry","name":"context-aware prompt building with workspace and tool registry","description":"CowAgent implements a Prompt Builder that dynamically constructs LLM prompts by combining system instructions, available tools, memory context, and workspace state. The builder maintains a tool registry that lists all callable tools with their signatures and descriptions, and injects this registry into prompts so the LLM knows what tools are available. The workspace tracks agent state (e.g., current directory, open files) and includes it in prompts for context-aware tool invocation. This enables the LLM to make informed decisions about which tools to use based on current context.","intents":["Automatically include available tools in prompts so the agent knows what it can do","Provide workspace context (e.g., current directory, open files) to inform tool selection","Dynamically adjust prompts based on memory and conversation history"],"best_for":["Developers building agents that need to be aware of available tools and workspace state","Teams that want to avoid manually specifying tool lists in prompts"],"limitations":["Prompt size grows with number of tools; large tool registries can exceed LLM context limits","Tool descriptions must be manually written; poor descriptions lead to tool misuse","Workspace state tracking is not automatic; must be explicitly updated by tool implementations","No optimization for prompt length; may waste tokens on redundant context"],"requires":["Tool registry with properly formatted tool definitions","System prompt template that includes placeholders for tools and context","Workspace state management (e.g., current directory tracking)"],"input_types":["tool definitions","workspace state","memory context","system instructions"],"output_types":["constructed prompts ready for LLM invocation"],"categories":["planning-reasoning","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"github-zhayujie--cowagent__cap_9","uri":"capability://tool.use.integration.browser.automation.and.terminal.command.execution","name":"browser automation and terminal command execution","description":"CowAgent provides built-in tools for browser automation (via Selenium or similar) and terminal command execution, enabling the agent to interact with web applications and execute system commands. These tools are registered in the tool registry and can be invoked by the agent during task planning. Browser automation allows the agent to navigate websites, fill forms, and extract data. Terminal execution allows the agent to run scripts, install packages, and perform system administration tasks. Both tools include safety constraints (e.g., command whitelisting, timeout limits) to prevent abuse.","intents":["Enable the agent to automate web-based tasks (e.g., form filling, data extraction)","Allow the agent to execute system commands and scripts as part of task planning","Build autonomous agents that can interact with external systems without human intervention"],"best_for":["Teams building autonomous agents for web automation and system administration","Developers who want to delegate repetitive web-based tasks to an AI agent"],"limitations":["Browser automation is slow (~1-5 seconds per action); not suitable for real-time interaction","Terminal command execution is dangerous; requires careful whitelisting and sandboxing to prevent privilege escalation","No built-in error recovery; failed commands may leave the system in an inconsistent state","Browser automation requires headless browser installation (e.g., Chrome, Firefox); adds deployment complexity"],"requires":["Selenium WebDriver or similar browser automation library","Headless browser installation (Chrome, Firefox, or similar)","Proper security configuration (command whitelisting, sandboxing, timeout limits)"],"input_types":["text (commands or URLs)","structured task descriptions"],"output_types":["command output","extracted web data","screenshots"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":56,"verified":false,"data_access_risk":"high","permissions":["Python 3.7+","Platform-specific credentials (WeChat token, Feishu app ID, DingTalk webhook, etc.)","Network connectivity to target platforms","API key for at least one LLM provider (OpenAI, Claude, Gemini, DeepSeek, Qwen, etc.)","Tools/skills registered in the agent's tool registry","Sufficient LLM context window to hold task description + available tools + execution history","Docker and docker-compose installed","Docker image built from Dockerfile (or pulled from registry)","Environment variables configured for LLM providers and channels","Persistent volumes for SQLite and skills directory (if using stateful deployment)"],"failure_modes":["Channel-specific features (e.g., WeChat group mentions, Feishu card formatting) require custom plugin handlers","Message ordering guarantees only within a single channel; cross-channel consistency requires application-level coordination","Daemon thread model adds ~50-200ms latency per channel due to thread context switching","Planning quality depends on LLM reasoning capability; weaker models (e.g., GPT-3.5) may fail on complex multi-step tasks","No built-in cost control; unbounded tool invocations can lead to high API bills or long execution times","Agent execution is synchronous; long-running tasks block the message handler thread","No formal verification of plan correctness; agent may execute harmful commands if not properly constrained","Docker deployment requires Docker and docker-compose installation; adds operational complexity","Containerized agents cannot access host system resources (e.g., terminal commands) without explicit volume mounts or privileged mode","Cloud deployment via LinkAI requires vendor lock-in; migrating away requires re-deployment","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.8406467874274438,"quality":0.5,"ecosystem":0.6000000000000001,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.28,"freshness":0.12}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:22.064Z","last_scraped_at":"2026-05-03T13:57:01.479Z","last_commit":"2026-05-03T13:50:11Z"},"community":{"stars":43982,"forks":10037,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=zhayujie--cowagent","compare_url":"https://unfragile.ai/compare?artifact=zhayujie--cowagent"}},"signature":"TJU0268ZlkNc/GWela6qoyeIkVsvxa2PnrrnIIzHi3jdUPLd+TRE2C4mJARinJwLFmXNezp3snFKCBPmDamKDA==","signedAt":"2026-06-20T22:47:59.912Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/zhayujie--cowagent","artifact":"https://unfragile.ai/zhayujie--cowagent","verify":"https://unfragile.ai/api/v1/verify?slug=zhayujie--cowagent","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}