plan-execute
Skill1 starsChanged 4 months agoMITView on GitHub
Implements code changes from a plan markdown file with verification. Use when user says 'execute this plan', 'implement the plan', 'follow this plan', 'run plan-execute', or provides a .md plan file to implement. Reads plan files (standalone or index linking sub-files), verifies each item exists before fixing, logs false positives to ignored.md. Use exclusively for executing existing plans, not creating them.
7 other projects use a version of this.
Adopted by
Most recently taken up by
- Esi-Abolfazl/skillstheir own version, first seen Nov 1, 2026
- allemaar/open-skillstheir own version, first seen Oct 13, 2026
- serbanghita/plan-critique-skillstheir own version, first seen Sep 7, 2026
- vanenshi/skillstheir own version, first seen Aug 5, 2026
- erhwenkuo/stock_skillstheir own version, first seen Apr 21, 2026
- okikusan-public/stock_skillstheir own version, first seen Apr 17, 2026
Other versions7 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 7
- okikusan-public/stock_skills★ 134same name
- longranger2/claude-gpt-workflow★ 76same name
- allemaar/open-skills★ 15same name
- serbanghita/plan-critique-skills★ 13same name
- erhwenkuo/stock_skills★ 3same name
- Esi-Abolfazl/skills★ 3same name
- vanenshi/skills★ 2same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository24 other files it gives its agents.
- app-hardenSkill
- arch-auditSkill
- arch-conformanceSkill
- ballot-advisorSkill
- bug-huntSkill
- canon-c50Skill
- canon-r6-mark-iiSkill
- code-scrutinySkill
- davinci-resolveSkill
- desktop-ui-craftSkill
- fsharp-upliftSkill
- pattern-architectSkill
- pine-script-proSkill
- plan-evalSkill
- plan-featureSkill
- refactor-proSkill
- rust-ci-greenSkill
- rust-commitSkill
- rust-test-separateSkill
- rust-upliftSkill
- rust-warn-fixSkill
- skill-forgeSkill
- skill-refineSkill
- sprintSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 10 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 6 edited versions we read keep these.
10
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
If no path is provided, ask the user for the plan file path before proceeding.
- Distinctive: line 42
If the work list is empty after filtering, report "No actionable items" and stop.
- Distinctive: line 46
Before your first implementation, read 3-5 files in the same directories you will modify.
- Distinctive: line 58
Before writing any code, confirm the item is real:
- Distinctive: line 64
Verify the described problem is present: Read the actual code and classify:
- Distinctive: line 78
When an item is Confirmed or Variant:
- Distinctive: line 81
Follow the plan's approach: if one is specified. If the plan only describes the problem, design the solution yourself.
- Very specific: line 92
When an item is Not found, Already resolved, or Not applicable, append to ignored.md in the plan's directory.
- Distinctive: line 106
After all items are processed, report:
- Distinctive: line 115
Project idiom wins: If the project uses callbacks, use callbacks — even if the plan suggests promises.
- Distinctive: line 117
Preserve intent: When adapting a fix to project conventions, it must still solve the original problem.
“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Plan Execute
You execute code change plans with surgical precision. Every change is correct, minimal, and consistent with the project's own conventions.
ultrathink
Input
The plan file path is: $ARGUMENTS
If no path is provided, ask the user for the plan file path before proceeding.
Workflow
Stage 1: Ingest the Plan
- Read the plan file at the provided path. If the file does not exist, tell the user and stop. Record the plan's parent directory — all relative paths resolve from there.
- Extract items and follow links recursively. A plan file can contain inline items, links to sub-files, or both. Do all of the following:
- Extract any actionable items directly present in the file.
- Scan for markdown links to other
.mdfiles ([text](path/to/file.md)). Read each linked file, resolving paths relative to the linking file's directory. Repeat recursively for any links found in sub-files. If a linked file does not exist, warn the user and continue with the remaining files. - Track which files you have already read. Skip any file encountered a second time (prevents circular links).
Read
ignored.mdfrom the plan's directory, if it exists. Parse each entry. An ignored entry matches a plan item when they reference the same file path AND the same code element (function, class, variable, or section). For cross-cutting items with no specific file (e.g., "all usages offoo()"), match on description. Fuzzy wording differences do not matter; the identifying key is what matters.Re-verification rule: Items with status
Already resolvedmust be re-verified every run, because regressions happen. OnlyNot foundandNot applicableitems are skipped. If re-verification confirms the item is still resolved, leave the existing ignored.md entry in place — do not append a duplicate.Build a work list of every discrete item from the plan and sub-files. Each item captures:
- Description of the item
- Location (file path, function/class/section, code pattern — whatever the plan provides)
- Suggested fix (if the plan specifies one)
- Whether it matches a skippable ignored.md entry
If the work list is empty after filtering, report "No actionable items" and stop.
Stage 2: Study Project Conventions
Before your first implementation, read 3-5 files in the same directories you will modify. Internalize the project's naming, formatting, error handling, and structural patterns. Check for .editorconfig, linter configs, and any CLAUDE.md / CONTRIBUTING.md style guides.
Priority order when conventions conflict: Documented guides (CLAUDE.md, CONTRIBUTING.md, linter configs) take precedence over patterns inferred from code. Project conventions take precedence over language-level defaults.
Stage 3: Verify and Implement
Process each item in plan order. After implementing any change, treat subsequent items as operating against the current file state — re-read and re-verify, do not rely on stale line numbers or prior assumptions.
3a. Locate and Verify
Re-read target files fresh for each item — do not rely on previously cached content.
Before writing any code, confirm the item is real:
- Locate the code. Use the file path, function name, class name, or code pattern from the plan — whichever is available. If the plan gives only a description with no location, use Grep and Glob to search the codebase for the described pattern. If the code cannot be found after a reasonable search, classify as
Not found. - Match by content, not line numbers. Line numbers shift. Locate the target code by matching function signatures, variable names, code patterns, or surrounding context. Use plan line numbers only as a starting hint, never as the sole locator.
Verify the described problem is present. Read the actual code and classify:
Classification Condition Action Confirmed The problem exists as described Implement the fix Variant A related but different problem exists at the same location Fix the actual variant; note the discrepancy Not found The code or the described problem does not exist Log to ignored.md Already resolved The problem was fixed since the plan was written Log to ignored.md Not applicable The code exists but the described problem is absent (guard clause already present, type system prevents it, etc.) Log to ignored.md When the call is ambiguous: If the code works but violates conventions or readability expectations noted by the plan author, treat it as Confirmed — trust the plan author's judgment. Only classify as Not applicable when the code is correct AND follows project conventions AND the plan's concern does not apply.
3b. Implement
When an item is Confirmed or Variant:
- Read surrounding context — understand the function, its callers, and its role before changing anything.
- Follow the plan's approach if one is specified. If the plan only describes the problem, design the solution yourself.
- Quality bar:
- Handles edge cases the plan may not have mentioned
- Follows project conventions from Stage 2
- Changes only what is needed for this item
- Introduces no new issues (security, broken imports, type errors)
- If a change causes a downstream failure (broken imports, type errors, test failures), resolve it before moving to the next item. If the failure cannot be resolved without exceeding the item's scope, revert the change and record it in the Errors section of the summary.
- Briefly state what was changed and why.
3c. Log to ignored.md
When an item is Not found, Already resolved, or Not applicable, append to ignored.md in the plan's directory. Create the file with a # Ignored Plan Items heading if it does not exist.
Format:
## [Brief description from the plan]
- **Plan source**: [filename and section/item reference]
- **Status**: Not found | Already resolved | Not applicable
- **Reason**: [Specific explanation referencing the actual code state you observed]Append only — do not remove existing entries. Before appending, check whether an entry with the same file path, code element, and status already exists. If so, skip the append to avoid duplicates.
Stage 4: Summary
After all items are processed, report:
- Implemented: count and brief list of changes made
- Ignored: count and brief list with reasons
- Errors: anything that could not be resolved
Rules
- Verify every item against actual code before implementation.
- Implement exactly what the plan specifies. Limit each change to the scope defined by its plan item.
- Project idiom wins. If the project uses callbacks, use callbacks — even if the plan suggests promises. Match the project's naming, structure, and patterns.
- Log every skip to ignored.md with a specific reason referencing what you found in the code.
- Preserve intent. When adapting a fix to project conventions, it must still solve the original problem.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

