Skip to content
Get packs

Guide · Agent skill

How to use the Convex Component Suggester agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The convex-suggest skill is a set of instructions that teaches an AI coding agent to notice when your code rebuilds something a Convex component already provides, such as cron jobs, rate limiting or file storage. This guide explains what the skill does, how to install it from a ZIP file, how to trigger it with natural requests and what it needs to run safely. By the end you will be able to add the skill to your agent and know what its suggestions look like.

Convex Component Suggesteropenclaw/clawhub · MIT

Quick start

  1. 1Download the ZIP. Download the convex-suggest skill ZIP from this page.
  2. 2Unzip the skill folder. Unzip the file so you have a convex-suggest folder containing SKILL.md.
  3. 3Place the folder. Move convex-suggest into ~/.claude/skills/ for all projects, or into your project's .claude/skills/ for one project.
  4. 4Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
  5. 5Work on a Convex project. Ask your agent for normal Convex work such as scheduled jobs, counters, uploads or rate limits.
  6. 6Review the suggestion. After the task, read the one suggestion, which names a component and quotes the triggering code.
  7. 7Answer yes or no. Reply yes to install the component, or no to drop the suggestion. Nothing is installed without your consent.

What the convex-suggest skill makes the agent do

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when the skill is relevant. This skill covers Convex, a backend platform. While the agent works on your task, it watches for code or requests that duplicate a Convex component, such as setInterval-based scheduling (cron), shared counter increments, or email sending.

The skill is passive. It does not interrupt your work. After the task you asked for is finished, the agent offers one suggestion. That suggestion names the component, quotes the specific code or phrase that triggered it, and explains why the component fits better. The skill says it never installs anything without your explicit yes, and if you decline or ignore the suggestion, the agent drops it and does not repeat it.

  • Matches include: crons, sharded-counter, rate-limiter, storage, search, presence, workflow, RAG and prosemirror-sync.
  • Only one suggestion at a time, picking the highest-priority match.
  • Does not fire on generic coding questions unrelated to Convex, such as sorting arrays or writing CSS.
  • Does not suggest a component you have already installed.

Install the convex-suggest skill

Download the skill ZIP from this page and unzip it. The skill folder is named convex-suggest and contains SKILL.md. Put it in one of these locations depending on how you want to use it:

For all your projects, place the convex-suggest folder inside ~/.claude/skills/. For one project only, place it inside that project's .claude/skills/ folder. If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/convex-suggest/
  • Project install: .claude/skills/convex-suggest/ inside your project folder
  • Claude apps: Settings → Capabilities → Skills, then upload the skill
~/.claude/skills/convex-suggest/SKILL.md

Trigger the skill with natural requests

You do not need a special command to use the skill. It is meant to run in the background while you code. Ask for normal Convex work, such as building a feature that sends emails, schedules a recurring job or limits how often users can submit requests. If the agent sees code that rebuilds a Convex component's job, it will offer a suggestion after finishing.

The skill's suggestion quotes your own code back to you, so you can judge whether it applies. For example, the source gives this pattern: a mutation that does post.likes + 1 while many users call it concurrently, which the agent may flag as a shared counter problem.

  • Example request: "Add a mutation that increments the like count on a post."
  • Example request: "Write a function that runs every hour to clean up old sessions."
  • Example request: "Let users upload profile photos and store them."
  • Example request: "Show who is currently typing in the chat."

Know what the skill needs and its safety limits

The skill requires a Convex project, because its suggestions only target Convex components such as @convex-dev/crons, @convex-dev/rate-limiter and @convex-dev/workflow. Installing a Convex component is optional and happens only with your explicit consent.

According to the skill's listing, it does not run scripts and does not use the network. Its only file is SKILL.md. Because the skill asks for consent before any install, it is not designed for users who want an agent to change dependencies automatically. Review every suggested change before you approve it.

  • Requires: a Convex project
  • Runs scripts: no
  • Uses network: no
  • Files: SKILL.md
  • Licence: MIT

Tips for getting useful suggestions

Suggestions are based on code and phrasing the agent can see, so keep the relevant Convex functions in the conversation or the project. If a suggestion does not fit, reply no and move on; the skill will not repeat it.

If you want the agent to check existing code for patterns that scale poorly, you can ask it to review a file. The skill is also useful for teams reviewing Convex code for things like concurrent counter updates.

  • Keep Convex code in the project so the agent can match patterns.
  • Answer yes or no; a no ends that suggestion.
  • Use a review request when you want a deliberate check rather than passive suggestions.

What people use it for

  • Spotting a hand-written setInterval job and suggesting the Convex crons component.
  • Flagging a mutation that increments a shared counter under heavy concurrent use and suggesting the sharded-counter component.
  • Noticing a hand-rolled rate limit kept in database counters and suggesting the rate-limiter component.
  • Catching a long-running multi-step action and suggesting the workflow component.
  • Suggesting Convex file storage in place of custom S3 or fs.write uploads.

Frequently asked questions

Is the convex-suggest skill free?

The skill is published under the MIT licence, according to its listing. Convex components suggested by the skill may be installed only if you say yes.

What do I need to use the convex-suggest skill?

You need a Convex project, since the suggestions target Convex components. The skill runs no scripts and uses no network, and it does not install anything without your consent.

Where do I put the skill so my agent can find it?

Place the convex-suggest folder in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project. In the Claude apps you can upload it under Settings → Capabilities → Skills.

Will the skill install packages on its own?

No. The skill is written to suggest a component, explain why, and wait for your explicit yes before any install. If you say no or ignore it, the suggestion is dropped.

Can I use this skill with a project that does not use Convex?

The skill is not meant for that. Its suggestions only target Convex components, so in non-Convex projects the agent may waste time on irrelevant matches.

How is this different from just asking an agent for a feature?

A normal request gets you code for the feature. This skill adds a passive check that, after the task, points out when a Convex component already provides the same pattern and explains why it may fit better.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers building on Convex who want an agent to point out when hand-built code duplicates a Convex component, and teams reviewing Convex code for patterns that scale poorly. It is a poor fit for projects that do not use Convex, or for anyone who wants an agent to change dependencies automatically without asking.

Download the skill