ClaudeSuperPower
SubmitLog in

improve-convex-plugin

Skill

Send THIS coding session's transcript (Claude Code or Codex) to the Convex quickstart backend (anteater) for an AI post-mortem that improves the whole system — the runbook, bootstrap script, and skills. Use after a quickstart run (success or failure), or whenever the user wants to \"send feedback\",

0Backend

Install

git clone https://github.com/get-convex/convex-backend-skill.git ~/.claude/skills/improve-convex-plugin

What is improve-convex-plugin?

Send THIS coding session's transcript (Claude Code or Codex) to the Convex quickstart backend (anteater) for an AI post-mortem that improves the whole system — the runbook, bootstrap script, and skills. Use after a quickstart run (success or failure), or whenever the user wants to \"send feedback\", \"report how that went\", or \"help improve the quickstart\".

What this can do

Capabilities declared in this component's own frontmatter — not inferred.

Inherit all session tools

Declares no tool restrictions — inherits every session tool

~91 tokens of context used while enabled, before you invoke anything

Documentation

README · ~2 min read

improve-convex-plugin

This skill ships the current session transcript to anteater's POST /review endpoint, which runs Claude over it and returns structured findings (ambiguous or wrong instructions, places the agent got stuck or repeated work, tooling failures, plus what worked) targeted at the runbook / bootstrap / skill / component. It is a system-improvement loop, not end-user feature feedback.

Consent — ask once, then remember

Sharing a session is opt-in, so you ask once and never nag again. Run the helper (below). If it prints CONSENT_REQUIRED and exits non-zero, the user has not yet chosen — stop and ask them, verbatim intent:

Reviews

Log in to leave a review.

No reviews yet — be the first.

Explore related

Other things in this space — across every part of the ecosystem, not just skills.