AgentSkillsCN

working-with-context

在分析现有项目或代码库时使用——将项目接入 Serena MCP 存储上下文,然后使用上下文感知分析,提及现有代码、模块和模式以获得更好的建议。

SKILL.md
--- frontmatter
name: working-with-context
description: Use when analyzing existing projects or codebases - onboard project to store context in Serena MCP, then use context-aware analysis that mentions existing code, modules, and patterns for better recommendations

Working with Context

Overview

Context management enables project-aware analysis that references existing code structure, tech stack, and patterns.

Workflow

1. Onboard project (one-time):

bash
shannon onboard /path/to/project --project-id myapp

Scans and stores:

  • File inventory and structure
  • Tech stack detection
  • Module identification
  • Pattern recognition

2. Use context in analysis:

bash
shannon analyze new_feature.md --project myapp

Difference:

  • Without context: Generic "add authentication" advice
  • With context: "Integrate with existing UserModel in models/user.py, use JWT pattern from auth.py"

Commands

shannon onboard PATH:

  • Scans codebase
  • Stores in ~/.shannon/projects/
  • Saves to Serena MCP

shannon context status:

  • Shows onboarded projects
  • Last update time

shannon context update:

  • Incremental update after code changes
  • Faster than full re-onboarding

shannon context clean:

  • Remove stale entries
  • Use --all for all projects

Storage

Context stored in:

  • Local: ~/.shannon/projects/<project-id>/
  • Serena MCP: Entities and relations
  • Persistent across sessions

When to Update

Update context after:

  • Major code changes (new modules)
  • Architecture changes
  • Dependency updates
  • Tech stack changes

Performance

  • Onboarding: <30s for most projects
  • Context loading: <1s (automatic in analyze)
  • Update: <10s (incremental)