Add documentation for web_fetch and web_search tools to the default
agent profile templates. Also adds additional options info for glob
and usage guidelines for the web tools.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* chore(deps): add yaml package for skill parsing
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat(agent): add skills system
Implement a skills system inspired by moltbot's approach:
- Skills are markdown files (SKILL.md) with YAML frontmatter
- Multi-source loading with precedence: bundled < user < workspace
- Eligibility filtering based on platform, binaries, and env vars
- Skills are automatically included in agent system prompt
- New AgentOptions: enableSkills, skillsBaseDir, extraSkillDirs
Includes two bundled skills:
- commit: Git commit helper with conventional commit guidelines
- code-review: Code review checklist and best practices
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor(skills): use profile-based skills instead of workspace
Change skill loading from workspace-based (.skills/) to profile-based:
- Skills now load from ~/.super-multica/agent-profiles/<profileId>/skills/
- Remove workspace and user skill sources
- Simplify to only bundled and profile sources
- Profile skills have higher precedence than bundled
This is more appropriate for non-coding agents where skills are
associated with agent identity rather than working directory.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* chore(deps): add web tools dependencies
Add undici, @mozilla/readability, linkedom, and turndown for
web fetching and HTML content extraction capabilities.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat(agent): add web_fetch and web_search tools
Port network tools from moltbot with the following features:
web_fetch:
- Fetch and extract readable content from URLs
- Dual extraction mode: readability (smart) or turndown (full page)
- SSRF protection (blocks private IPs, localhost)
- Response caching with TTL
- Redirect handling
web_search:
- Brave Search API for traditional results
- Perplexity API for AI-synthesized answers
- Region/language settings
- Freshness filtering (Brave only)
- Auto-detects provider from available API keys
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat(agent): register web_fetch and web_search tools
Add web tools to the agent's tool registry.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Add documentation for glob tool, exec auto-backgrounding, and process
actions. Reorganize into File Operations and Command Execution sections
with usage guidelines.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Add a glob tool that allows agents to find files matching glob patterns.
Features:
- Supports standard glob patterns (e.g., **/*.ts, src/**/*.{js,jsx})
- Results sorted by modification time (most recent first)
- Configurable result limit and ignore patterns
- Default ignores for node_modules, .git, dist, etc.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
* feat(agent): add auto-backgrounding to exec tool
- Add yieldMs parameter to exec tool (default 5s) - commands that don't
complete within this time automatically run in background
- Create shared process-registry.ts for unified process management
- Refactor process.ts to use shared registry
- Add --debug CLI flag for session message logging
- Signal isolation: backgrounded processes ignore abort signals
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(session): preserve tool_use/tool_result pairs during compaction
Previously, session compaction simply kept the last N messages, which
could break tool_use/tool_result pairs if the cut point fell between
them. This caused "tool_call_id is not found" errors from the API.
Now compaction finds a safe cut point that starts from either:
- A user message without tool_result
- An assistant message whose tool_use is needed by the next tool_result
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat(session): use Kimi as default model for summary compaction
- Auto-detect MOONSHOT_API_KEY from environment
- Use moonshot-v1-128k (cheaper than k2-thinking)
- Fall back to tokens mode if API key not available
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* docs: add rule to never use git commit --amend
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* docs: clarify git amend rule for immediate fixes
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Translate all comments, JSDoc strings, and documentation text from Chinese to English across 19 files for improved accessibility and consistency with the broader developer community.
Co-authored-by: Claude Haiku 4.5 <noreply@anthropic.com>
* feat(agent): add context window guard to prevent token overflow
Implement token-aware context management that validates context window
size on agent initialization and provides intelligent message compaction
based on actual token usage rather than simple message count.
Key changes:
- Add context-window module with guard, token estimation, and types
- Support both "count" (legacy) and "tokens" (new default) compaction modes
- Warn when context window < 32K tokens, block when < 16K tokens
- Trigger compaction at 80% utilization, target 50% after compaction
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat(agent): add summary-based compaction using LLM
Implement intelligent compaction that uses LLM to generate summaries
of older messages instead of simply truncating them. This preserves
important context like key decisions, TODOs, and technical details.
Key changes:
- Add summarization.ts with compactMessagesWithSummary functions
- Support chunked summarization for very large histories
- Add "summary" compaction mode alongside "count" and "tokens"
- Auto-resolve API key from environment based on provider
- Graceful fallback to "tokens" mode if model/apiKey unavailable
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Introduce a new interactive CLI that provides a REPL interface for continuous conversation with the agent. Features include multi-line input mode, session management, built-in commands, and streaming output integration with existing agent infrastructure.
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
- Add UUIDv7 session ID examples and continuation workflow
- Add new Sessions section documenting storage format
- Add --thinking CLI option example
- Document process tool output buffering and auto-cleanup
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Document agent profile system and profile CLI commands
- Document exec and process tools with usage examples
- Update project structure to reflect new directories
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add TTL-based auto-cleanup for terminated processes (1 hour)
- Add 'cleanup' action for manual removal of terminated processes
- Fix buffer overflow when single chunk exceeds MAX_OUTPUT_BUFFER
- Track terminatedAt timestamp for cleanup eligibility
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Agent Profile module for managing agent identity, soul, tools,
memory, and bootstrap configuration
- Add profile CLI (pnpm agent:profile) for creating/listing/showing profiles
- Default sessionId to UUIDv7 instead of "default"
- Expose Agent.sessionId as public readonly property
- Improve exec/process tools error handling (no more crashes on spawn errors)
- Add 'output' action to process tool for reading stdout/stderr
- Better tool descriptions to guide agent in choosing exec vs process
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add pnpm workspace and Turborepo configuration
- Extract Gateway SDK to packages/sdk as independent package
- Add Next.js + shadcn Web client in apps/web
- Update root package.json with turbo scripts
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Implement Hub that auto-connects to Gateway, manages agents via REST API,
and routes WebSocket messages to agents by payload.agentId with echo responses
sent back to the original sender.
- Add Hub with GatewayClient auto-connect, agent CRUD, and message routing
- Add Console (NestJS) with REST API and static pages (management + demo client)
- Switch Gateway registration from explicit event to query-based on connect
- Remove deprecated types (RegisterPayload, metadata, SendMessagePayload)
- Add @nestjs/serve-static for serving console UI
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Dockerfile and build script for Gateway containerization
- Add USAGE.md documenting SDK usage for Client/Agent implementations
- Move socket.io-client to production dependencies (required by SDK)
- Remove redundant timestamp field from RoutedMessage (UUID v7 contains timestamp)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
