Analyze Request (Standalone)

This is the standalone analysis skill (/ccx:analyze). You interact directly with the user.

Note: When used inside /ccx:run, the pipeline's Adaptive Plan phase (Phase 1) takes precedence. That version runs as a subagent (ccx:planner) and combines analysis + planning in one pass.

Steps

1. Call mcp__ccx__load_project_context with the current project directory.
2. Call mcp__ccx__get_session to check for previous context.
3. Index first: Call mcp__ccx__trigger_index(project_dir) to discover all scopes and identify stale/new ones.
4. For each relevant scope, call mcp__ccx__get_scope_with_children(project_dir, scope) to load cached analysis with hierarchy.
   • Fresh → use cached analysis, skip reading code.
   • Stale → re-analyze only changed files, then save via mcp__ccx__save_analysis_cache with file_hashes, children, parent.
   • New (uncached) → full analysis, then save via mcp__ccx__save_analysis_cache with file_hashes, children, parent.
   • Scope naming: project-root-relative file path, no extension, lowercase (e.g. "src/ccx/mcp_server").
5. Analyze the request against the project context.

Rules

• Do NOT assume anything that is ambiguous. Use AskUserQuestion to clarify (with options, 2-4 items).
• Do NOT reference any code. Work only with the user's intent.
• If the request is clear enough, proceed without questions.
• Keep intent to ONE sentence.
• Scope should be module/layer/feature level, not file level.

Output

Present the analysis to the user in this format:

Analysis Result

• Intent: [one sentence summary]
• Scope: [list of affected modules/layers/features]
• Constraints: [any constraints, including relevant exception rules]

If there were ambiguities resolved via questions, note what was clarified.

The user's request is: $ARGUMENTS