How deekshitha-palla/week-5-project works
Code Scout is a conversational coding agent: a user asks for help, the agent routes the request through an LLM/tool loop, and returns a response, while persisting session history. Keyword-matched skills inject task-specific guidance and can unlock optional tools. The tool suite covers repository exploration, sandboxed file and command operations, planning, web research, paper lookup, and remote MCP servers. The excerpts establish several internal calls; other tool dispatch links are supported by the SOURCE INDEX, so details of unsampled implementations are not inferred.
Main entry points: agent.py Agent Loop · paper.py Paper Search & Read · web.py Web Search & Fetch · skill_matcher.py Skill Matcher
Components 16
Agent Runtime
- Agent Loop
agent.py - Session History
agent.py - Todo Plans
tools/plan.py
Skill Routing
- Skill Loader
skill_loader.py - Skill Matcher
skill_matcher.py
Codebase Tools
- Search & Outline
tools/explore.py - Workspace Files
tools/file.py - Command Runner
tools/execute.py
Research Integrations
- Web Search & Fetch
tools/web.py - Paper Search & Read
tools/paper.py - MCP Connector
mcp_connector.py
Other components
- User
- MCP Servers
- LLM API
- Web Services
- Hugging Face Papers API
Connections 18
- User → Agent Loop: sends prompt
agent.py - Agent Loop → LLM API: requests completion
agent.py - Agent Loop → Session History: loads and saves
agent.py - Agent Loop → Skill Loader: loads skills
agent.py - Agent Loop → Skill Matcher: matches request
agent.py - Skill Matcher → Skill Loader: uses skill types
skill_matcher.py - Agent Loop → Todo Plans: checks pending
agent.py - Agent Loop → Search & Outline: dispatches tools
agent.py - Agent Loop → Workspace Files: dispatches tools
agent.py - Agent Loop → Command Runner: dispatches tools
agent.py - Agent Loop → Web Search & Fetch: dispatches tools
agent.py - Agent Loop → Paper Search & Read: dispatches tools
agent.py - Agent Loop → MCP Connector: connects and dispatches
agent.py - MCP Connector → MCP Servers: calls remote tools
mcp_connector.py - Paper Search & Read → Hugging Face Papers API: requests paper data
tools/paper.py - Paper Search & Read → Web Search & Fetch: fetches fallback page
tools/paper.py - Web Search & Fetch → Web Services: searches and fetches
tools/web.py - Agent Loop → User: returns response
agent.py