
execute-plan
Execute an approved implementation plan exactly and safely. Use for ExecPlan-style plan execution, resumable phase checklists, multi-hour implementation work, migrations, significant refactors, and plans that require verification checkpoints, status tracking, and final reporting.
- Installs
- 558
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/buiducnhat/cobrew --skill execute-plan- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads execute-plan automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Execute Plan
Overview
Execute a pre-approved plan with strict adherence to scope, sequence, and verification. Treat the plan as a living document that records progress, findings, decisions, and outcomes as execution proceeds.
The input is typically: execute-plan docs/.plans/YYMMDD-HHmm-<plan-slug>/SUMMARY.md
Do not redesign the plan during execution. If ambiguity or blockers appear, stop and ask.
Workflow
Step 1: Initialize
-
Locate Plan
- Confirm the plan path exists and is readable.
- If a directory is provided, locate
SUMMARY.mdinside it.
-
Load Execution Context
- Load only the project context relevant to the plan:
- If
docs/SUMMARY.mdexists, read it first. - Load only task-relevant detail docs.
- Prioritize
Code Standarddocs for implementation conventions. - If docs conflict with code or user intent, use the available input/question before broad changes.
- If
- Review the plan’s phase files, context/orientation notes, dependencies, acceptance criteria, idempotence/recovery notes, decision log, and open questions.
- If an older plan is missing living-plan sections (
Progress,Surprises & Discoveries,Decision Log,Outcomes & Retrospective), add those sections before execution when the plan is otherwise clear.
- Load only the project context relevant to the plan:
-
Select Execution Mode (Explicit Rule)
- Default mode: Batch
- Use Interactive when any of the following is true:
- High-risk changes (auth, payments, migrations, security-critical logic)
- Irreversible operations (data migrations, destructive scripts)
- Unclear acceptance criteria
- User explicitly requests checkpoints
- If mode is unclear, ask once and proceed with user choice.
-
Find Next Pending Phase
- First
[ ]phase - If none, first
[-]phase - If no pending/in-progress phases remain, go to final verification.
- First
-
Critical Plan Sanity Check
- Ensure each phase has:
- clear objective
- file targets
- verification commands
- observable acceptance criteria
- If observable acceptance criteria are missing but can be inferred from explicit phase verification and exit criteria, add them to the phase before execution.
- If essential details are missing or contradictory, stop and request clarification.
- Ensure each phase has:
Step 2: Execute Per-Phase Loop
For each phase in order:
-
Skip Completed
- If status is
[x], continue to next phase.
- If status is
-
Mark In Progress
- Update phase status to
[-]before making changes. - Add a timestamped
Progressentry inSUMMARY.mddescribing the phase start.
- Update phase status to
-
Execute Exactly
- Implement only the tasks defined in that phase.
- Do not expand scope without approval.
- Write the minimum code that satisfies the phase. No speculative features, no abstractions for single-use code, no error handling for impossible scenarios. See Simplicity first and Surgical changes rules below.
- If implementation reveals a material surprise, record it under
Surprises & DiscoveriesinSUMMARY.md. - If a decision is needed to stay within scope, record the decision and rationale under
Decision Logbefore continuing.
-
Verify Phase
- Run the phase-specific verification commands from the plan.
- At minimum, run relevant tests/checks tied to touched files.
- Record concise verification evidence in
SUMMARY.mdunderProgressor the phase's requested evidence location.
-
Handle Failures
- If verification fails:
- Attempt focused fixes within phase scope.
- Re-run verification.
- If still failing or root cause is outside scope, stop and report blocker.
- If verification fails:
-
Mark Complete
- Update phase status to
[x]only after verification passes. - Add a timestamped
Progressentry with the verification result and changed files for that phase.
- Update phase status to
-
Progress Report
- Interactive mode: report and wait for confirmation before next phase.
- Batch mode: report briefly and continue immediately.
Step 3: Final Verification
After all phases are complete:
-
Project-Wide Validation
- Run full lint/type-check suite
- Run all relevant tests (or full test suite if required by the plan)
- Run build verification if applicable
-
Stabilize
- Fix regressions introduced during execution.
- Re-run failed checks until green or blocked.
-
Manual Validation Checkpoint
- If user/manual QA is required, ask explicitly and pause:
Verifiedto accept- or provide feedback for follow-up iteration
- If user/manual QA is required, ask explicitly and pause:
-
Update Outcomes
- Complete
Outcomes & RetrospectiveinSUMMARY.mdwith final result, verification summary, approved deviations, and follow-ups.
- Complete
Step 4: Completion Artifacts
-
Documentation Sync
- If behavior/architecture/codebase expectations changed, update the
docsartifacts.
- If behavior/architecture/codebase expectations changed, update the
-
Create Execution Report
- File:
docs/.plans/YYMMDD-HHmm-<plan-slug>/EXECUTION-REPORT.md - Include all required sections below.
- File:
-
Prepare Final Gate
- Do not archive before final user confirmation. The user may choose
Need verify, and execution must continue against the same plan path.
- Do not archive before final user confirmation. The user may choose
Step 5: Final Confirmation Gate
After completion artifacts are done, ask the user for a final confirmation using the input/question tool with exactly these options:
Confirm: End sessionNeed verify
Handle the selected option as follows:
-
Confirm: End session- Archive the plan folder to
docs/.plans/archived/. - Announce:
Execution complete. Report archived at docs/.plans/archived/YYMMDD-HHmm-<plan-slug>/EXECUTION-REPORT.md. - End the execution session.
- Archive the plan folder to
-
Need verify- Allow the user to provide verification feedback/details.
- Do not archive.
- Continue the execution loop to address feedback, then re-run verification and completion steps as needed.
Execution Report Standard
EXECUTION-REPORT.md must use the following template: references/execution-report-template.md
Rules
- Respect project standards: follow
docs/and related project docs. - Follow the plan strictly: no silent scope changes.
- Stop on blocker: missing dependency, contradictory instructions, or unexplained failures.
- No guessing: ask for clarification when uncertain.
- Verify before complete: never mark phase done without passing checks.
- Keep the plan alive: update progress, discoveries, decisions, and outcomes in
SUMMARY.mdas execution proceeds. Do not rely on chat history for important execution state. - Idempotency: prefer safe/re-runnable operations.
- Simplicity first: Implement the minimum code that satisfies the phase's exit criteria. No features beyond what the plan asks for. No abstractions for single-use code. No configurability that wasn't requested. If you write 200 lines and it could be 50, rewrite it.
- Surgical changes: Touch only what the phase requires. Don't "improve" adjacent code, comments, or formatting. Don't refactor things that aren't broken. Match existing style even if you'd do it differently. Only remove imports/variables/functions that your changes orphaned — don't delete pre-existing dead code unless the plan asks for it. Every changed line should trace to a phase task.
- Write artifacts in language same with the current session.
FAQ
What does the execute-plan skill do?
Execute an approved implementation plan exactly and safely. Use for ExecPlan-style plan execution, resumable phase checklists, multi-hour implementation work, migrations, significant refactors, and plans that require verification checkpoints, status tracking, and final reporting.
How do I install execute-plan?
Run npx skills add https://github.com/buiducnhat/cobrew --skill execute-plan in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/execute-plan/ (personal) or .claude/skills/execute-plan/ (project).
Which AI agents can use execute-plan?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is execute-plan safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains execute-plan?
It is published by buiducnhat in buiducnhat/cobrew. It has 558 recorded installs.
More from buiducnhat/cobrew
brainstorm
buiducnhat/cobrew
Collaborative discovery and design framing for ambiguous or high-risk work. Use when requirements are unclear, multiple approaches are possible, or you need to turn an idea into a validated design brief before planning or coding.
write-plan
buiducnhat/cobrew
Create detailed, execution-ready implementation plans for complex or high-risk changes without coding. Use for ExecPlan-style work, multi-hour changes, significant refactors, migrations, resumable phase checklists, and work that should be handed off to execute-plan with clear validation.
docs
buiducnhat/cobrew
Create or refresh project documentation in docs/. Use when a project needs docs initialized or existing docs synced to the current codebase.
Popular code quality & workflow skills
grill-me
mattpocock/skills
A relentless interview to sharpen a plan or design.
improve-codebase-architecture
mattpocock/skills
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
handoff
mattpocock/skills
Compact the current conversation into a handoff document for another agent to pick up.
domain-modeling
mattpocock/skills
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
codebase-design
mattpocock/skills
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
mattpocock/skills
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.