Agent Skills: Plan Tracking Protocol

Plan tracking protocol (portable). Progress tracking with checkboxes, iteration workflow, completion/blocker signals. Use when executing multi-step plans, tracking task progress, or working from plan files. IMPORTANT - Always update the plan file after completing tasks.

UncategorizedID: tianjianjiang/smith/smith-plan

Install this agent skill to your local

pnpm dlx add-skill https://github.com/tianjianjiang/smith/tree/HEAD/smith-plan

Skill Files

Browse the full folder contents for smith-plan.

Download Skill

Loading file tree…

smith-plan/SKILL.md

Skill Metadata

Name
smith-plan
Description
Plan tracking protocol (portable). Progress tracking with checkboxes, iteration workflow, completion/blocker signals. Use when executing multi-step plans, tracking task progress, or working from plan files. IMPORTANT - Always update the plan file after completing tasks.

Plan Tracking Protocol

Portable protocol for tracking plan progress across iterations. Platform-agnostic — works with any AI agent that can read/write files.

Load if: Executing plans, tracking multi-step tasks Prerequisites: @smith-ctx/SKILL.md

CRITICAL: Plan Sync Protocol

After completing ANY task, you MUST update the plan file.

This ensures the next iteration sees your progress.

  • Sync the plan file right after completing work — do not defer it
  • Mark tasks complete as soon as they are finished
  • Keep plan format changes compatible with checkbox parsing

Progress Tracking Format

Use this format in plan files for trackable progress:

## Tasks

- [x] Task 1: Set up project structure
- [x] Task 2: Create database schema
- [ ] Task 3: Implement API endpoints <- CURRENT
- [ ] Task 4: Add authentication
- [ ] Task 5: Write tests

## Progress Log

### Iteration 1 («TIMESTAMP»)
- Completed: Task 1, Task 2
- Notes: Used PostgreSQL instead of MySQL per user preference

### Iteration 2 («TIMESTAMP»)
- Working on: Task 3
- Blockers: None

Workflow Per Iteration

1. LOAD: Read plan file
   -> Locate and read the plan

2. IDENTIFY: Find first uncompleted task (- [ ])
   -> Skip tasks marked [x]

3. EXECUTE: Complete the current task
   -> Do the actual implementation work

4. SYNC: Update plan file with progress
   -> Mark task [x], add progress log entry
   -> Write to SAME file path

5. CONTINUE or COMPLETE
   -> If more tasks: proceed to next
   -> If all done: signal completion

Plan Update Template

When updating the plan, use this pattern:

# 1. Mark task complete (change - [ ] to - [x])
# 2. Add progress log entry with timestamp
# 3. Note any blockers or changes
# 4. Save to the same file path

Completion-Aware Reload

An auto-loaded or auto-listed plan is stale, not live, when its checkboxes are all - [x] (or git log / a session memory shows the work merged). Disregard a stale plan — report it and ask — rather than resuming it. Verify checkbox state before treating any reloaded plan as work to continue.

Completion Signal

When all tasks are done, output this exact phrase:

PLAN COMPLETE: All tasks finished successfully.

Blocker Signal

If you hit a blocker requiring human input:

BLOCKER: «description of issue requiring human decision»

Important Rules

  1. ALWAYS update the plan file after completing work
  2. NEVER skip the sync step - next iteration depends on it
  3. Use checkbox format - [ ] / - [x] for trackable tasks
  4. Add timestamps to progress log entries
  5. Note file changes in progress log for traceability

Related

  • @smith-ctx/SKILL.md - Context management
  • @smith-plan-claude/SKILL.md - Claude Code automation (hooks, scripts)
  • @smith-ralph/SKILL.md - Ralph Loop iterative development

Before You Finish

Per iteration:

  1. Read plan file
  2. Find first - [ ] task
  3. Execute the task
  4. Mark - [x], add progress log entry
  5. Continue or signal completion