AgentSkillsCN

planning-with-files

为复杂任务实施 Manus 式的基于文件的规划。创建 task_plan.md、findings.md 与 progress.md 文件。当用户开始复杂的多步骤任务、研究项目,或任何需要超过 5 次工具调用的任务时,可使用此技能。现在还新增了 /clear 后的自动会话恢复功能,以及可选的 Git Worktree 模式。

SKILL.md
--- frontmatter
name: planning-with-files
version: "3.2.0"
description: Implements Manus-style file-based planning for complex tasks. Creates task_plan.md, findings.md, and progress.md. Use when starting complex multi-step tasks, research projects, or any task requiring >5 tool calls. Now with automatic session recovery after /clear and optional Git worktree mode.
user-invocable: true
allowed-tools:
  - Read
  - Write
  - Edit
  - Bash
  - Glob
  - Grep
  - WebFetch
  - WebSearch
hooks:
  PreToolUse:
    # Location verification for worktree mode (Write/Edit operations only)
    - matcher: "Write|Edit"
      hooks:
        - type: command
          command: |
            SCRIPT_DIR="${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/plugins/planning-with-files}/scripts"
            # Check for worktree config and verify location
            if [ -f ".planning-config.json" ]; then
              # Determine which verification script to use
              if command -v pwsh &> /dev/null && [[ "$OSTYPE" == "msys" || "$OSTYPE" == "win32" || "$OS" == "Windows_NT" ]]; then
                pwsh -ExecutionPolicy Bypass -File "$SCRIPT_DIR/verify-location.ps1" 2>/dev/null || powershell -ExecutionPolicy Bypass -File "$SCRIPT_DIR/verify-location.ps1" 2>/dev/null || bash "$SCRIPT_DIR/verify-location.sh"
              else
                bash "$SCRIPT_DIR/verify-location.sh"
              fi
              # Exit code 1 means wrong location - block the operation
              if [ $? -eq 1 ]; then
                exit 1
              fi
            fi
    # Show task_plan context for all operations (except in worktree mode for read/grep/glob)
    - matcher: "Write|Edit|Bash|Read|Glob|Grep"
      hooks:
        - type: command
          command: |
            # In worktree mode, silence read/grep/glob operations
            if [ -f ".planning-config.json" ]; then
              MODE=$(jq -r '.mode' .planning-config.json 2>/dev/null || echo "")
              if [ "$MODE" = "worktree" ]; then
                # Silent for read/grep/glob in worktree mode
                case "$CLAUDE_TOOL_NAME" in
                  Read|Glob|Grep) exit 0 ;;
                esac
              fi
            fi
            # Show task_plan for write/edit/bash or standard mode
            cat task_plan.md 2>/dev/null | head -30 || true
  PostToolUse:
    - matcher: "Write|Edit"
      hooks:
        - type: command
          command: "echo '[planning-with-files] File updated. If this completes a phase, update task_plan.md status.'"
  Stop:
    - hooks:
        - type: command
          command: |
            SCRIPT_DIR="${CLAUDE_PLUGIN_ROOT:-$HOME/.claude/plugins/planning-with-files}/scripts"
            if command -v pwsh &> /dev/null && [[ "$OSTYPE" == "msys" || "$OSTYPE" == "win32" || "$OS" == "Windows_NT" ]]; then
              pwsh -ExecutionPolicy Bypass -File "$SCRIPT_DIR/check-complete.ps1" 2>/dev/null || powershell -ExecutionPolicy Bypass -File "$SCRIPT_DIR/check-complete.ps1" 2>/dev/null || bash "$SCRIPT_DIR/check-complete.sh"
            else
              bash "$SCRIPT_DIR/check-complete.sh"
            fi

Planning with Files

Work like Manus: Use persistent markdown files as your "working memory on disk."

FIRST: Check for Previous Session (v2.2.0)

Before starting work, check for unsynced context from a previous session:

bash
# Linux/macOS
uv run python ${CLAUDE_PLUGIN_ROOT}/scripts/session-catchup.py "$(pwd)"
powershell
# Windows PowerShell
& (Get-Command python -ErrorAction SilentlyContinue).Source "$env:USERPROFILE\.claude\skills\planning-with-files\scripts\session-catchup.py" (Get-Location)

If catchup report shows unsynced context:

  1. Run git diff --stat to see actual code changes
  2. Read current planning files
  3. Update planning files based on catchup + git diff
  4. Then proceed with task

Important: Where Files Go

  • Templates are in ${CLAUDE_PLUGIN_ROOT}/templates/
  • Your planning files go in your project directory
LocationWhat Goes There
Skill directory (${CLAUDE_PLUGIN_ROOT}/)Templates, scripts, reference docs
Your project directorytask_plan.md, findings.md, progress.md

Quick Start

Standard Mode

Before ANY complex task:

  1. Create task_plan.md — Use templates/task_plan.md as reference
  2. Create findings.md — Use templates/findings.md as reference
  3. Create progress.md — Use templates/progress.md as reference
  4. Re-read plan before decisions — Refreshes goals in attention window
  5. Update after each phase — Mark complete, log errors

Worktree Mode (Multi-Task Parallel Development)

For parallel multi-task development with isolated Git worktrees:

  1. Start worktree mode — Use /planning-with-files:worktree [task-name] [target-branch]

    • Example: /planning-with-files:worktree feature-auth main
    • Creates a new Git worktree directory (.worktree/feature-auth/)
    • Creates a task branch with planning files inside the worktree
    • Main directory stays on original branch (no switching!)
    • Multiple worktrees can exist simultaneously for parallel tasks
  2. Navigate to worktreecd .worktree/feature-auth

    • Work on your task in this isolated environment
    • Follow standard planning workflow
  3. Complete and merge — Use /planning-with-files:complete [target-branch] from inside the worktree

    • Deletes planning files from worktree
    • Navigates to root directory
    • Merges task branch to target
    • Removes the worktree directory
    • Deletes the task branch

Multi-Task Example:

bash
# Start task 1
/planning-with-files:worktree fix-auth-bug
cd .worktree/fix-auth-bug

# In another terminal, start task 2 (parallel!)
/planning-with-files:worktree refactor-api
cd .worktree/refactor-api

# Each task has its own directory and branch
# No conflicts, no branch switching needed

Benefits:

  • Work on multiple tasks simultaneously without conflicts
  • Each task has its own isolated environment
  • No need to switch branches in the main directory
  • Easy cleanup when tasks are complete

Note: Planning files go in your project root, not the skill installation folder.

The Core Pattern

code
Context Window = RAM (volatile, limited)
Filesystem = Disk (persistent, unlimited)

→ Anything important gets written to disk.

File Purposes

FilePurposeWhen to Update
task_plan.mdPhases, progress, decisionsAfter each phase
findings.mdResearch, discoveriesAfter ANY discovery
progress.mdSession log, test resultsThroughout session

Critical Rules

1. Create Plan First

Never start a complex task without task_plan.md. Non-negotiable.

2. The 2-Action Rule

"After every 2 view/browser/search operations, IMMEDIATELY save key findings to text files."

This prevents visual/multimodal information from being lost.

3. Read Before Decide

Before major decisions, read the plan file. This keeps goals in your attention window.

4. Update After Act

After completing any phase:

  • Mark phase status: in_progresscomplete
  • Log any errors encountered
  • Note files created/modified

5. Log ALL Errors

Every error goes in the plan file. This builds knowledge and prevents repetition.

markdown
## Errors Encountered
| Error | Attempt | Resolution |
|-------|---------|------------|
| FileNotFoundError | 1 | Created default config |
| API timeout | 2 | Added retry logic |

6. Never Repeat Failures

code
if action_failed:
    next_action != same_action

Track what you tried. Mutate the approach.

The 3-Strike Error Protocol

code
ATTEMPT 1: Diagnose & Fix
  → Read error carefully
  → Identify root cause
  → Apply targeted fix

ATTEMPT 2: Alternative Approach
  → Same error? Try different method
  → Different tool? Different library?
  → NEVER repeat exact same failing action

ATTEMPT 3: Broader Rethink
  → Question assumptions
  → Search for solutions
  → Consider updating the plan

AFTER 3 FAILURES: Escalate to User
  → Explain what you tried
  → Share the specific error
  → Ask for guidance

Read vs Write Decision Matrix

SituationActionReason
Just wrote a fileDON'T readContent still in context
Viewed image/PDFWrite findings NOWMultimodal → text before lost
Browser returned dataWrite to fileScreenshots don't persist
Starting new phaseRead plan/findingsRe-orient if context stale
Error occurredRead relevant fileNeed current state to fix
Resuming after gapRead all planning filesRecover state

The 5-Question Reboot Test

If you can answer these, your context management is solid:

QuestionAnswer Source
Where am I?Current phase in task_plan.md
Where am I going?Remaining phases
What's the goal?Goal statement in plan
What have I learned?findings.md
What have I done?progress.md

When to Use This Pattern

Use for:

  • Multi-step tasks (3+ steps)
  • Research tasks
  • Building/creating projects
  • Tasks spanning many tool calls
  • Anything requiring organization

Skip for:

  • Simple questions
  • Single-file edits
  • Quick lookups

Templates

Copy these templates to start:

Scripts

Helper scripts for automation:

Standard Mode Scripts

  • scripts/init-session.sh — Initialize all planning files
  • scripts/check-complete.sh — Verify all phases complete
  • scripts/session-catchup.py — Recover context from previous session (v2.2.0)

Worktree Mode Scripts (v2.7.2)

  • scripts/worktree-init.sh — Start a new worktree session (bash)
  • scripts/worktree-init.ps1 — Start a new worktree session (PowerShell)
  • scripts/worktree-complete.sh — Complete and merge worktree (bash)
  • scripts/worktree-complete.ps1 — Complete and merge worktree (PowerShell)

Advanced Topics

Anti-Patterns

Don'tDo Instead
Use TodoWrite for persistenceCreate task_plan.md file
State goals once and forgetRe-read plan before decisions
Hide errors and retry silentlyLog errors to plan file
Stuff everything in contextStore large content in files
Start executing immediatelyCreate plan file FIRST
Repeat failed actionsTrack attempts, mutate approach
Create files in skill directoryCreate files in your project