Cheap Agent's MCP server, plan-writing skill, and /cheapag:status, /cheapag:allow, /cheapag:clean, and the opt-in edit guard (/cheapag:hook).

GitHub stars
335

Install

/plugin install cheapag@cheap-agent

Run it inside Claude Code after adding the marketplace with /plugin marketplace add muaaz9329/cheap-ag.

Contents

Discovered from source

Skills 7

!npx -y cheap-agent@1.4.3 allow $ARGUMENTS

The developer ran this consent command themselves; its output is above. Report it in one line.

Only because the user invoked this skill: If no output appears above (Codex): run that command yourself once, exactly as written, with $ARGUMENTS replaced by whatever the user typed after the skill name (or nothing). It needs network for npx; ask for that if the sandbox blocks it. Do not run cheapag allow again or change consent in any other way.

!npx -y cheap-agent@1.4.3 clean $ARGUMENTS

Report the result above to the user in one or two lines.

If no output appears above (Codex): run that command yourself once, exactly as written, with $ARGUMENTS replaced by whatever the user typed after the skill name (or nothing). It needs network for npx; ask for that if the sandbox blocks it.

!npx -y cheap-agent@1.4.3 hook $ARGUMENTS

The developer ran this command themselves; its output is above. Report it in one line.

Only because the user invoked this skill: If no output appears above (Codex): run that command yourself once, exactly as written, with $ARGUMENTS replaced by whatever the user typed after the skill name (or nothing). It needs network for npx; ask for that if the sandbox blocks it. Do not turn the edit guard on or off again or in any other way.

!npx -y cheap-agent@1.4.3 model --list

The developer ran this model command themselves; its output is above. Set the worker's models with them:

  1. Ask one question for the main model (it writes the code): the 4 suggestions as the options, and "Other" for any other model id from the provider's catalog.
  2. If the picked model lists efforts, ask one question for its effort, with default first.
  3. Ask one question for the fallback model (used on a rate limit): none first, then the suggestions. If it lists efforts, ask for its effort too.
  4. Run npx -y cheap-agent@1.4.3 model --main <id> [--effort <e>] [--fallback <id|none>] [--fallback-effort <e>] once, leaving out the flags for default efforts, and report its last line.

Always show the user this line, as is: For more models, exit Claude/Codex and run: npx cheap-agent model

Only because the user invoked this skill: If no output appears above (Codex): run npx -y cheap-agent@1.4.3 model --list yourself once, exactly as written. It needs network for npx; ask for that if the sandbox blocks it. Do not change the models in any other way.

Writing a Cheap Agent plan

The cheap worker sees the whole plan, then implements the steps you name in delegate. It works in an isolated copy of the repo and does only what a step says, so each step must stand on its own: exact files, what to build, and a check that proves it works.

Save plans under .plan/ in the repository, e.g. .plan/rate-limit.md (the developer may read, edit or commit them). Pass the path relative to the repository root.

Grammar

# <plan title>                              ← required, exactly one H1
<shared context, optional, ≤ ~15 lines>
## <id> <short title>                       ← id = integer or integer+letter (3, 3a; s3, step-3 also work)
files: <path>, <dir/>, ...                  ← required; scope + overlap key
do: <what to build; continuation lines indented 2+ spaces>
done-when: <condition>; <condition>         ← required; `backticked` = executed
dont: <constraint>; <constraint>
needs: <step id>, ...
difficulty: easy | medium | hard            ← picks the worker model; wins over delegate's
kind: fix | change                          ← fix: a bug to reproduce; change: no reproduction expected
model: <model>
deps: <package@range>                       ← the step needs a new package
  • Keys are lowercase, at the start of the line, never bullets (- files: is wrong). Continuation lines are indented 2+ spaces.
  • In done-when, every backticked fragment is run as a shell command in the worker's copy and must exit 0. Put only commands in backticks; plain text is the worker's responsibility.
  • Unknown keys (e.g. api:, example:) are passed to the worker as hints.

Rules

  • Size: one step ≈ under an hour of junior-developer work. Split anything bigger.
  • One change per step: a do: that lists three or more numbered items is refused at delegate; split it into 3a, 3b, … and order them with needs:.
  • Details as bullets: write a do: as - bullets or prose; numbered or lettered items count as separate changes, and three or more get the task refused. do: 1. extract X 2. move Y 3. update the import is refused; - extract X, - move Y, - update the import is accepted.
  • Checks: give every step an executable done-when (a test command, a type check), not only prose. Name the one test file or test id the step concerns, never the whole suite.
  • Kind: say kind: fix for a bug to reproduce or kind: change when no reproduction is expected; otherwise the worker infers it from the step's wording.
  • Source: put the user's request or the issue verbatim in the shared context, or pass it as delegate(source=…) (text or a file path): the worker checks your steps against it and reports where they disagree.
  • Scope: files: lists every file or directory the step may touch, accurately. It also decides which steps can run in parallel.
  • Constraints: add dont: for the mistakes you expect (new dependencies, touching auth, changing public APIs).
  • Grouping: make independent steps separate comma groups so they run in parallel; use needs: when one step must land first.
  • Packages: declare a new package with deps:; the worker cannot install packages itself.
  • Share context once at the top (conventions, existing helpers to reuse) instead of repeating it per step.

Example

# Rate limiting for auth endpoints
Redis is already wired in src/lib/redis.ts. Keep middleware style consistent with src/middleware/*.

## 1 limiter lib
files: src/lib/rateLimit.ts, test/rateLimit.test.ts
do: sliding-window limiter `limit(key, max, windowSec) -> {allowed, retryAfter}` using redis ZSET.
done-when: `npm test -- rateLimit` passes; tests cover allow, block, window expiry
dont: add dependencies

## 2 wire into login
files: src/api/login.ts, src/middleware/, test/login.test.ts
do: 5 req/min per IP on POST /api/login; on block return 429 + Retry-After.
done-when: `npm test -- login` passes; new test: 6th request → 429
dont: touch auth/session logic
needs: 1

## 3 docs
files: docs/api.md
do: document the 429 response and Retry-After.
done-when: section "Rate limits" exists under /api/login

Delegating it

  • steps="3": one step, one job. steps="3-5": a range, run in order in one job. steps="1-2,3": comma-separated groups, run in parallel unless their files: overlap or needs: orders them.
  • delegate(plan=".plan/rate-limit.md", steps="1-2,3")
  • delegate(plan=".plan/rate-limit.md", steps="2", note="use the existing ipFrom(req) helper"): a note is at most 300 characters, for one-off guidance.

!npx -y cheap-agent@1.4.3 provider $ARGUMENTS

The developer ran this provider command themselves; its output is above.

  • If it listed the providers set up (no provider was given), ask the user one question: which provider to use, with the listed ones as the options (the one in use marked). Then run npx -y cheap-agent@1.4.3 provider <id> once and report its last lines.
  • If it switched or failed, report it in one or two lines, including any consent line and fix it printed.

A provider that is not set up needs a key: tell the user to exit and run npx cheap-agent login in a terminal. Never ask for or handle a key.

Only because the user invoked this skill: If no output appears above (Codex): run npx -y cheap-agent@1.4.3 provider $ARGUMENTS yourself once, exactly as written, with $ARGUMENTS replaced by whatever the user typed after the skill name (or nothing). It needs network for npx; ask for that if the sandbox blocks it. Do not change the provider or consent in any other way.

!npx -y cheap-agent@1.4.3 status

Show the status above to the user as is. Add a line only if something needs their action (a BLOCKED job to reply to, a missing consent, a Go window near its cap).

If no output appears above (Codex): run that command yourself once, exactly as written, with $ARGUMENTS replaced by whatever the user typed after the skill name (or nothing). It needs network for npx; ask for that if the sandbox blocks it.

MCP servers 1

Sponsors

Details

Marketplace
muaaz9329/cheap-ag
Source
plugin
Category
Community