| name | architecture-context |
| description | One-shot architecture tracing — get entry points, call graph neighborhood, and source snippets for any natural-language question about how code works. |
| allowed-tools | ["mcp__mycelium__context","mycelium_context"] |
| category | navigation |
| icon | 🏗️ |
| marketplace_examples | [{"query":"How does request routing work?","tool":"mcp__mycelium__context"},{"query":"Trace handle_request to get_user","tool":"mcp__mycelium__context"},{"query":"Show how auth reaches token verification","tool":"mcp__mycelium__context"}] |
architecture-context — one-shot architecture tracing
This Skill exposes the mycelium_context tool (MCP) and mycelium context
CLI command: a single call that accepts a natural-language task description and
returns the most relevant entry-point symbols, their call-graph neighborhood,
and source location spans.
Use this before chaining lower-level tools like mycelium_search_symbol,
mycelium_get_callers, and mycelium_get_callees. One mycelium_context call
typically saves 5–20 round-trips.
When to invoke this Skill
Use when:
- The user asks "how does X work?" or "trace A to B".
- You need a broad orientation in an unfamiliar area of the codebase.
- You want entry points + neighbors + source spans in a single call.
Do NOT use when:
- You already know the exact symbol path and only need its direct callers/callees
→ use the
call-graph Skill.
- You want a precise multi-hop graph query → use the
hyphae-query Skill.
Tool reference
MCP: mycelium_context
task (string) Natural-language architecture question or "trace A to B".
max_nodes (int?) Maximum graph nodes returned. Default: 30.
max_code_blocks (int?) Maximum source snippets. Default: 6.
output_format (string?) "json" (default), "text", or "msgpack".
CLI: mycelium context
mycelium context --task "trace ServeHTTP to HandlerFunc"
mycelium context --task "how does auth work" --format json
mycelium context --task "function:calls(#AuthService)" --max-nodes 50
--task TEXT Natural-language task or Hyphae selector (required)
--root PATH Project root (default: current directory)
--max-nodes INT Maximum graph nodes (default: 30, max: 100)
--max-code-blocks INT Maximum source snippets (default: 6, max: 25)
--format json|text Output format (default: json)
Both surfaces return entry_points, nodes, edges, code_blocks, stats,
and agent_summary. The agent_summary.next_step field tells you whether to
proceed from the returned context or chain a narrower tool.
