catlog22/Claude-Code-Workflow
1
0
Fork 0
mirror of https://github.com/catlog22/Claude-Code-Workflow.git synced 2026-03-03 15:43:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(planning-agent): remove skip-CLI logic, keep Prior Analysis as context

Browse Source 
Revert the Phase 2 skip when no explorations. CLI execution should always
run but use Prior Analysis block from task description as primary context
when exploration files are absent.
This commit is contained in:
catlog22
2026-03-02 16:22:04 +08:00
parent e54d76f7be
commit 4f3ef5cba8
1 changed files with 4 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
.claude/agents/cli-lite-planning-agent.md
View File

@@ -122,24 +122,20 @@ const planObject = generatePlanFromSchema(schema, context)
## Execution Flow
**Prior Analysis Fast Path**: When prompt contains "No exploration files" or exploration context is empty, SKIP Phase 2 (CLI execution). Instead, generate plan directly from the task description's `## Prior Analysis` block + schema. This saves tool budget for file writing (Phase 4). Go: Phase 1 → Phase 3 (direct generation) → Phase 4 → Phase 5.
```
Phase 1: Schema & Context Loading
├─ Read schema reference (plan-overview-base-schema or plan-overview-fix-schema)
├─ Aggregate multi-angle context (explorations or diagnoses)
├─ **Check**: Has exploration files?
│ ├─ YES → Continue to Phase 2
│ └─ NO (Prior Analysis) → Skip Phase 2, generate plan directly in Phase 3
├─ If no explorations: use "## Prior Analysis" block from task description as primary context
└─ Determine output structure from schema
Phase 2: CLI Execution (SKIP when no explorations)
├─ Construct CLI command with planning template
Phase 2: CLI Execution
├─ Construct CLI command with planning template (include Prior Analysis context when no explorations)
├─ Execute Gemini (fallback: Qwen → degraded mode)
└─ Timeout: 60 minutes
Phase 3: Parsing & Enhancement
├─ (Normal) Parse CLI output sections OR (Prior Analysis) Generate tasks directly from task description
├─ Parse CLI output sections
├─ Validate and enhance task objects
└─ Infer missing fields from context