Skip to content
Agentmart
Submit a tool

mestack

Kevin's harness-agnostic engineering dispatcher.

Skill by kevherro

IndexedMIT
Write a review

Chalk Labs

Need help deploying mestack?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/kevherro-mestack

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Kevin's harness-agnostic engineering dispatcher. Picks a playbook, applies mestack principles, and keeps subagent fan-out in the parent. Use for /mestack, mestack-mode, or non-trivial engineering work (bug, feature, refactor, investigation, overnight run).

Use cases: agent-building

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Nov 1, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/mestack

What's inside

Skill name
mestack
Description
Kevin's harness-agnostic engineering dispatcher. Picks a playbook, applies mestack principles, and keeps subagent fan-out in the parent. Use for /mestack, mestack-mode, or non-trivial engineering work (bug, feature, refactor, investigation, overnight run).
License
MIT
Instructions
140 lines in SKILL.md · view on GitHub

Files (10)

  • SKILL.md
  • build.md · playbooks/
  • fix.md · playbooks/
  • investigate.md · playbooks/
  • leave-running.md · playbooks/
  • reshape.md · playbooks/
  • amp.md · references/
  • grok.md · references/
  • harness.md · references/
  • models.md · references/

SKILL.md (first lines)

# mestack

Read [references/harness.md](references/harness.md) once per task to
resolve available capabilities and fallbacks before choosing tools.

## Start

1. Open a todo list. First item: read the Principles index below.
2. Match a playbook. Open its file. Copy its steps into the todo list
   verbatim, before any task-specific todos. A skipped step stays in
   the list with `skip: <reason>`.
3. Name each principle that changes a decision, and the choice it
   changed. Cite only leaves you read this session. A citation with no
   decision is a miss.

If no playbook fits, run the **figure-it-out** skill. Do not invent
a second dispatcher.

## Playbooks

| When | File |
| --- | --- |
| Read-only question: how, why, are we sure | [playbooks/investigate.md](playbooks/investigate.md) |
| Defect to reproduce and fix | [playbooks/fix.md](playbooks/fix.md) |
| New or changed behavior | [playbooks/build.md](playbooks/build.md) |
| Behavior-preserving restructure | [playbooks/reshape.md](playbooks/reshape.md) |
| User is leaving; drive to a finish condition | [playbooks/leave-running.md](playbooks/leave-running.md) |

## Principles

Read the leaf in full when you apply it.

- **Smallest change** (`principle-smallest-change`). Sizing a diff, or
  tempted to add a layer. Delete and shrink first.
- **Named shape** (`principle-named-shape`). Before writing logic: the
  data, the invariant, the structure that makes the rest obvious.
- **Prove on the artifact** (`principle-prove-on-the-artifact`). After
  the work, before done. Run the real thing. The **prove** skill owns
  how.
- **Root cause** (`principle-root-cause`). Debugging. Reproduce first.
  Do not silence the crash.
- **Attack the premise** (`principle-attack-the-premise`). Two fixes
  on one belief failed the same check. Test the belief, not a third
  fix.
- **Test behavior** (`principle-test-behavior`). Writing or keeping a
  test. Fails when behavior breaks, survives a refactor.
- **Explain the number** (`principle-explain-the-number`). Before
  reporting a measurement. Name the limiter, rule out other readings.
- **Parent owns fan-out** (`principle-parent-owns-fan-out`). Any
  parallel work. Children do not spawn children.
- **Reversible without asking** (`principle-reversible-without-asking`).
  Tempted to ask permission for work you can undo. Do it, show the
  result.
- **Encode the rule** (`principle-encode-the-rule`). About to write the
  same instruction twice. Put it in a check, not more prose.
- **One home** (`principle-one-home`). A fact already lives somewhere.
  Point at it. Do not copy it.
- **Boundary discipline** (`principle-boundary-discipline`). Wiring
  CLI, config, network, or adapters. Validate at the edge.
- **Build the lever** (`principle-build-the-lever`). Non-trivial
  work. The artifact is a rerunnable tool.
- **Exhaust the design space** (`principle-exhaust-the-design-space`).
  No precedent. Compare two or three real alternatives.
- **Experience first** (`principle-experience-first`). Product or
  API tradeoff. Consumer result over a cheaper implementation.
- **Guard the context window** (`principle-guard-the-context-window`).
  Large reads or fan-out. Parent keeps pointers, not dumps.
- **Make operations idempotent** (`principle-make-operations-idempotent`).
  Commands, startup, loops. Retries converge.
- **Migrate then delete** (`principle-migrate-callers-then-delete-legacy-apis`).
  Replacing an internal API. One wave, no dual path.
- **Minimize reader load** (`principle-minimize-reader-load`). Hard
  to trace. Cut layers and hidden state.
- **Outcome-oriented execution** (`principle-outcome-oriented-execution`).
  Planned rewrite. Converge on the target, not throwaway middles.
- **Redesign from first principles**
  (`principle-redesign-from-first-principles`). New requirement.
  Integrate as if it had always been there.
- **Separate before serializing**
  (`principle-separate-before-serializing-shared-state`). Concurrent
  writers. Split ownership before adding a lock.
- **Sequence verifiable units** (`principle-sequence-verifiable-units`).
  Sweeps, migrations, commit order. Each unit ends in a check.
- **Type system discipline** (`principle-type-system-discipline`).
  Types and signatures. Illegal states must not compile.

## Routing

- Nontrivial subsystem you do not yet understand: the **how** skill.
- After implementation or a claimed fix: the **prove** skill.
- No project `verify-*` skill, and the repo has a runnable app:
  **create-verification-skill**. Prove offers this once; do not
  generate it as a side effect of a fix.
- Feature map or verify skill drifted: **maintain-verification-skill**.
- Parallel coverage or competing attempts: parent spawns, unique output
  paths, then the parent reads and synthesizes. See harness.md.
- Large or cross-cutting run, even if a small version would be fix or
  build: **figure-it-out**.
- Concurrency, retries, state machines, parsers, or unsafe code where
  tests cannot reach the bad case: **create-formal-verification-skill**
  (once per repo), then that project `formal-*` skill. Drifted:
  **maintain-formal-verification-skill**.
- Measured a speedup or regression: **benchmark-checklist** before
  reporting it.
- The user corrected the same mistake twice: **correct**.
- Adversarial review of a change: **interrogate**.
- Catch-up on recent work: **recall**. Catch-up within this chat:
  **ketchup**. Teach-me: **teach**. Explicit TDD: **tdd**.
- Opening a PR, committing, pushing, or contacting anyone: only when
  the user asked for that action.

## Autonomy

Proceed on reversible work. Pause for force-push to shared branches,
deploys, data deletion, and messages to other people.

"Don't stop", "going to bed", "run until done": the leave-running
playbook. Log via **show-me-your-work**.

From kevherro/mestack, MIT.

More from kevherro/mestack

Signals

Stars
0
Forks
2
Open issues
0
Last commit (this folder)
Oct 23, 2026 (11 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from kevherro

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get mestack running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying mestack?

CallGet help
Cite this page

mestack skill. Agent Mart. https://agentmart.dev/skills/kevherro-mestack (last updated 2026-10-24)