Execution Plan Review logo

Execution Plan Review

Community
richkuo
execution-plan-review

Use when the user wants to review or revise the ordering/model/effort/plan/plan-effort assignments on a milestone's GitHub issues — "review the execution plan", "/execution-plan-review", "show me the model assignments", or piecemeal revisions like "11 should be medium" or "plan 17 at xhigh". Renders the assignment table from the issues' Execution blocks, validates revisions, and writes them back. Stage 5 of the new-app-pipeline.

Overview

Publisherrichkuo
Repositoryrk-skills
Skill nameexecution-plan-review
Stars
50
Forks
8
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by richkuo on GitHub. Read the source before you install it.

Installation

Install the Execution Plan Review AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/richkuo/rk-skills/tree/main/skills/execution-plan-review
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/richkuo/rk-skills.git /tmp/rk-skills
mkdir -p .claude/skills
cp -r /tmp/rk-skills/skills/execution-plan-review .claude/skills/execution-plan-review
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Execution Plan Review in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Execution Plan Review on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Execution Plan Review is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

execution-plan-review

Show a milestone's per-issue ordering and execution assignments as one table, take the user's revisions, and sync the GitHub issues, which are the source of truth. milestoneplan is the read-only view of this table; it routes Execution-block fixes here.

Steps

1. Render the table from the issues, not from memory

Fetch every milestone issue (gh issue list --milestone "<title>" --json number,title,body, with the title quoted) and parse the ## Execution blocks (read a legacy fableplan first: line as plan first:, and write back only plan first:):

IssueCDepends onRuns afterBuild modelEffortValidateplan first?Plan

Cell rules:

  • Validate is the effective <model> · <effort> the pipeline dispatches. The model comes from the [C<score>] band (a missing prefix routes as band 3) unless a Validate model: line stamps Fable 5.1, Opus 5.5, or <Name> (Codex CLI[, <model-id>]); the effort is the band default unless a Validate effort: line stamps one. The validate-issue step 6 band table is the authority; do not restate it. Show a stamp as Fable 5.1 · low (stamped; band Opus 5.5 · high) or Astra · low (Codex CLI, gpt-6-astra; stamped; band Opus 5.5 · high), and a clamp as Opus 5.5 · high (stamped low → high: low is a Fable-only tier). Only low clamps on an Opus validate; medium runs as stamped. When the user gives a skip list (issue numbers, or "all", they plan to pass to milestone-workflow as skipValidate), a listed issue shows skipped (run arg) when no open in-milestone predecessor (Depends on or Runs after), missing [C..] prefix, or missing Execution block predicts a fallback; otherwise it keeps the route with (listed; validates: <reason>). Add one note line that the run checks the evidence, the edit history, and the base branch at run time, and validates any issue that fails a check. The skip list is a run arg: never write it into an issue.
  • Effort shows the tier the pipeline runs. A Claude build runs only at high, so a Claude row stamped any other tier shows the clamp, e.g. high (stamped xhigh → high: Claude builds run at high only), the same way milestoneplan shows it. A Codex CLI or Cursor CLI build shows its stamped tier.
  • Plan shows <model> · <effort> on plan first: Yes issues: the model defaults to Opus 5.5, and on an Opus 5.5 plan the effort defaults to xhigh when the score is 85 or higher (or the title has no score) and to high below 85; a Fable plan runs only at high; a stamped Plan model: or Plan effort: line shows its value with (stamped). On a No issue show — only when no line is stamped; any stamped line, a default value included, shows as <stamp> (inert — no plan stage runs).
  • An absent ordering field shows as missing, never none.

Then add 2–3 sentences on the pattern (dominant bands, which issues plan first at score ≥ 75, the review trigger) as a sanity check against the validate-issue step 6 band table.

2. Take revisions

  • Shorthand: "11 should be medium", "12 depends on 8 and 9", "13 runs after 12", "clear 14's dependencies", "plan 17 at medium", "validate 271 at medium". Resolve row-vs-issue ambiguity against the table just shown; confirm in half a sentence when still ambiguous. A bare effort revision means the build effort; confirm in half a sentence when the issue also has a plan stage.
  • External CLI builds ("build 275 with luna on codex at max", "277 grok on cursor high", "278 on cursor with cursor-grok-4.6-high") write - **Build model:** <Name> (Codex CLI) or <Name> (Cursor CLI), with , <model-id> inside the parenthetical when the user names an id, and the tier to Effort:. max is a Codex CLI-only tier: write a Cursor max as xhigh. A Claude build runs only at high, so write every other tier on a Claude build, Fable included, as high; say the clamp applied.
  • "validate at " writes - **Validate effort:** <tier>; "validate with fable" or "validate with fable at low" writes - **Validate model:** Fable 5.1 (or Opus 5.5) plus any tier; "validate 20 with astra on codex at low" writes - **Validate model:** Astra (Codex CLI, <model-id>) with the id the user names and - **Validate effort:** low (low to max); "plan at " writes - **Plan effort:** <tier>; "plan with fable" writes - **Plan model:** Fable 5.1, and the default Opus 5.5 plan needs no line (an Opus plan stamped low runs at high). A revision that names the band default for that stage removes the line instead of stamping it; say so. A milestone-wide revision ("every Fable validate at medium", "all plans at medium", "validate both with fable on low") applies, in one batch, to every issue whose stage runs on that model after the batch.
  • Push back once when a revision conflicts with the score band (validate-issue step 6 / prd-to-issues): plan below score 75, or dropping below Opus on a money/security/irreversible-deletion issue. Give one recommendation with the reason; the user decides.
  • Batch revisions; do not round-trip to GitHub per message.
  • A Plan effort revision on a plan first: No issue is inert. Say so once and drop it, or ask whether they meant to turn plan on. A flip to Yes plans on Opus 5.5 at the score default (xhigh at 85 or higher, high below) unless the same batch stamps a Plan model: or Plan effort: line. A flip Yes → No on an issue with a Plan model: or Plan effort: line strips every such line during write-back; say which lines you dropped. Offer once to strip any inert stamp already on an issue instead of silently carrying it forward.
  • Preserve the edge kind: Depends on stays a hard prerequisite, Runs after stays ordering-only. Never move an issue between the fields to simplify the graph.
  • Before writing: verify every referenced issue exists, reject self-references, deduplicate each list, reject a predecessor present in both fields, and recursively fetch referenced issues outside the milestone until the explicit ordering graph closes.
  • Reject the whole batch before write-back if the combined graph (every Depends on and Runs after edge, including unchanged and external issues) contains a cycle. Edit no issue, and show the cycle path.

3. Write back

Load github-issue-format. For each changed issue, gh issue edit preserving the entire body and changing only the intended Execution block lines; the footer verb flips to Updated. Strip \r from gh-fetched bodies. When edits collide with concurrent issue edits, re-fetch and re-apply only your delta. Re-run the graph validation after any concurrent-edit re-fetch and before retrying a write.

4. Confirm

Re-render the final table once after all revisions land, and say it is what the milestone workflow will execute.

Failure modes

SituationDo this
An issue lacks an Execution blockDerive model, effort, and plan first: from the [C..] band per prd-to-issues. Stamp no Validate effort:, Plan model:, or Plan effort: line unless the user asked for that model or tier. Flag the backfilled block in the table
An issue lacks one or both ordering fieldsBackfill from the approved prd-to-issues graph when available; otherwise infer from Approach/Problem, mark the value inferred in the table, and confirm before write-back
Revision references a row that does not existShow the table again; ask which issue they meant
Revision puts a Claude build's effort at any tier other than highWrite high and say the clamp applied: every Claude build, Fable 5.1 included, runs only at high. A Codex CLI or Cursor CLI build accepts low to xhigh as stamped, and Codex also accepts max
Revision names an external CLI build or validate without a known default model idOnly Luna (Codex CLI), Astra (Codex CLI), and Grok (Cursor CLI) resolve an id; before writing, ask for the id in half a sentence and write it inside the parenthetical
Revision puts a Fable 5.1 validate stage at xhighWrite it as stamped. The revision is the explicit ask that permits xhigh on Fable; the unstamped default stays high (the LLM Attribution Footer section of CLAUDE.md owns this rule)
Revision puts a Fable plan stage at any tier other than highRemove the Plan effort: line and say the clamp applied: a Fable plan, on Fable 5.1 or any later Fable version, runs only at high
Revision names a validate modelWrite - **Validate model:** Fable 5.1, Opus 5.5, or <Name> (Codex CLI[, <model-id>]); a stamped model overrides the band (validate-issue step 6). Reject a Cursor CLI validate and keep the prior stamp; say the pipeline runs a validate pass only in the Codex read-only file sandbox, and Cursor has no write boundary. Ask about any other Claude name in half a sentence
Revision names a validate effortWrite the Validate effort: line. A Fable validate (stamped) runs every tier low to xhigh as stamped; a Codex CLI validate runs low to max as stamped. An Opus validate (stamped, or unstamped at any score) runs medium, high, and xhigh as stamped and raises low to high; say so, and offer a Validate model: stamp as the way to reach a Fable or Codex validate
A stamped Build model or Effort diverges from the issue's current [C..] band defaultApply validate-issue/issue-editing.md (Edit the title): a stamp at or above the band default on a field (a heavier model, a Fable 5.1 build, a Codex CLI or Cursor CLI harness, a higher effort on a Codex CLI or Cursor CLI build, plan first: Yes) is a deliberate override from any session. Keep it, and mark it in the table as stamped above band. Restamp a stamp below the band default up to the default and report it, unless the user names the lower stamp deliberate in this session, with the step 2 push-back; a lower effort is deliberate only on a Codex CLI or Cursor CLI build. A Claude build Effort: stamp other than high is no override: show the clamp per the Effort cell rule, restamp it to high on write-back, and report it. A run-time rescore keeps a harness or Fable 5.1 stamp and only adds plan first: Yes
Revision flips plan first: Yes → No on a band-3 ([C75]+) issueWrite it and say the issue loses its plan stage. Build effort stays high
Revision names a plan effortWrite the Plan effort: line with low, medium, high, or xhigh. A revision to the issue's default tier (xhigh on an Opus plan at score 85 or higher or with no score, else high) removes the line. On a Fable plan, every tier runs at high (see the Fable plan row)
Revision names a plan modelWrite the Plan model: line with Fable 5.1. A revision to Opus 5.5 removes the line. A Plan model revision on a plan first: No issue is inert, the same as a Plan effort revision

Frequently asked questions

What does the Execution Plan Review AI skill do?

Use when the user wants to review or revise the ordering/model/effort/plan/plan-effort assignments on a milestone's GitHub issues — "review the execution plan", "/execution-plan-review", "show me the model assignments", or piecemeal revisions like "11 should be medium" or "plan 17 at xhigh". Renders the assignment table from the issues' Execution blocks, validates revisions, and writes them back. Stage 5 of the new-app-pipeline.

Why use Execution Plan Review on TypingMind?

Because you install it once and use it with any model. Execution Plan Review is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Execution Plan Review in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/richkuo/rk-skills/tree/main/skills/execution-plan-review. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Execution Plan Review?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Execution Plan Review?

As many as you like. As long as a model supports skills, you can use Execution Plan Review with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Execution Plan Review AI skill free?

Yes. It is published on GitHub by richkuo under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇