AgentSkillsCN

iterating-plan

在充分调研与更新的基础上，对现有实施计划进行迭代优化。在根据反馈更新现有计划、细化计划阶段或成功标准，或调整计划范围时使用。

SKILL.md
--- frontmatter
name: iterating-plan
description: Iterate on existing implementation plans with thorough research and updates. Use when updating existing plans based on feedback, refining plan phases or success criteria, or adjusting plan scope.

Iterating on Implementation Plans

Update existing implementation plans based on user feedback through a surgical, verified approach.

Workflow

  1. •Identify plan and feedback - Parse input for plan path and requested changes
  2. •Read and understand - Read the entire existing plan
  3. •Research if needed - Only for changes requiring new technical understanding
  4. •Confirm approach - Present understanding before modifying
  5. •Update surgically - Make precise edits, preserve good content
  6. •Sync and review - Commit changes, offer further iteration

Step 1: Handle Input

If NO plan file provided:

code
Which plan would you like to update? Please provide the path.
Tip: `ls -lt thoughts/shared/plans/ | head`

If plan file but NO feedback:

code
I've found the plan at [path]. What changes would you like to make?
Examples: "Add a phase for X", "Update success criteria", "Split Phase 2"

If BOTH provided: Proceed directly to Step 2.

Step 2: Read and Understand

  1. •Read the existing plan FULLY - No limit/offset parameters
  2. •Parse requested changes - What to add/modify/remove
  3. •Determine if research needed - Only for new technical understanding

Step 3: Research If Needed

Skip if changes are simple (reordering, rewording, scope adjustments).

Only spawn research if the feedback requires new technical understanding or validating assumptions.

  1. •Create a research todo list with TodoWrite (only if research is non-trivial)
  2. •Spawn parallel, focused sub-tasks (do not research serially if tasks are independent)
    • •Be explicit about exact directories to search and what signals to extract
    • •Ask sub-tasks to return file:line references for every claim

Code investigation (pick the right one):

  • •codebase-locator - Find relevant files quickly
  • •codebase-analyzer - Understand implementation details and constraints
  • •codebase-pattern-finder - Find similar patterns to model after

Historical context:

  • •thoughts-locator - Find related research or decisions
  • •thoughts-analyzer - Extract insights from prior documents
  1. •Read any newly-identified files FULLY into main context (Read without limit/offset)
  2. •Wait for ALL sub-tasks to complete before synthesizing
  3. •Verify sub-task results: if something seems off, spawn a follow-up task immediately

Step 4: Confirm Approach

code
Based on your feedback, I understand you want to:
- [Change 1]
- [Change 2]

My research found:
- [Relevant discovery]

I plan to update by:
1. [Specific modification]
2. [Another modification]

Does this align with your intent?

Get confirmation before proceeding.

Step 5: Update the Plan

  1. •Make surgical edits using the Edit tool
  2. •Maintain existing structure unless explicitly changing it
  3. •Keep file:line references accurate
  4. •Ensure consistency - new phases follow existing pattern
  5. •Update related sections - scope, success criteria, approach

Step 6: Sync and Review

bash
thoughtcabinet sync -m "Plan update: <description>"

Present changes made and offer further iteration.

code
I've updated the plan at `thoughts/shared/plans/[filename].md`

Changes made:
- [Specific change 1]
- [Specific change 2]

The updated plan now:
- [Key improvement]
- [Another improvement]

Would you like any further adjustments?

Guidelines

Be Skeptical:

  • •Question vague feedback - ask for clarification
  • •Verify technical feasibility with research
  • •Point out conflicts with existing phases

Be Surgical:

  • •Precise edits, not wholesale rewrites
  • •Only research what's necessary
  • •Preserve content that doesn't need changing

Be Interactive:

  • •Confirm understanding before changes
  • •Allow course corrections
  • •Don't disappear into research

Track Progress:

  • •Use TodoWrite to track research/update work if complex
  • •Mark research tasks complete as you finish them

No Open Questions:

  • •Research or ask immediately if uncertain
  • •Never update plan with unresolved questions

Success Criteria Structure

Maintain two-category structure when updating:

  1. •Automated Verification - Commands like make test, npm run lint
  2. •Manual Verification - UI/UX, performance, edge cases

Path Handling

  • •The thoughts/searchable/ directory contains hard links for searching
  • •Always use canonical paths when referencing or writing files
    • •Use: thoughts/shared/prs/123.md
    • •Not: thoughts/searchable/shared/prs/123.md