convex-improve-convex-plugin
get-convex/agent-skills
Send your coding session to Convex for AI-powered system improvements and quickstart refinement.
What is convex-improve-convex-plugin?
This skill captures your coding session transcript and sends it to the Convex team's anteater service for automated analysis. The review identifies patterns like ambiguous instructions, agent failures, and tooling gaps—helping improve the Convex quickstart system. Sharing is opt-in and user-controlled.
- Captures current coding session transcript via anteater helper script
- Performs AI post-mortem analysis identifying instruction ambiguities, agent-stuck patterns, and tooling failures
- Returns structured findings targeted at runbooks, bootstrap scripts, skills, and components
- Provides opt-in consent flow (Always / Just this once / Never) with persistent user choice
- Redacts secrets and tokens before upload to protect sensitive data
- Summarizes highest-severity findings with suggested fixes and system wins
How to install convex-improve-convex-plugin
npx skills add https://github.com/get-convex/agent-skills --skill convex-improve-convex-plugin- Convex coding session with Claude/Codex transcript (.jsonl file)
- Access to anteater endpoint (provided by Convex)
- User consent to share transcript (opt-in, remembered)
How to use convex-improve-convex-plugin
- 1.Run the anteater helper with your session idea: `curl -fsSL "<anteater>/send-transcript" | bash -s -- --idea "<one-line app idea>"`
- 2.If prompted with CONSENT_REQUIRED, ask the user to choose Always, Just this once, or Never
- 3.Re-run the command with `--consent always|once|never` if consent was needed
- 4.Watch for output markers: REVIEW_SOURCE (transcript found), REVIEW_SUBMITTED (accepted), REVIEW_DONE (findings ready)
- 5.Summarize the highest-severity findings for the user, focusing on system improvements not user data
Use cases
- Improve Convex quickstart documentation by sharing real session friction points
- Identify gaps in agent skills or bootstrap scripts through actual usage patterns
- Help Convex team debug why agents get stuck on common tasks
- Contribute anonymized session data to refine the agent-skill ecosystem
- Validate whether new Convex features are discoverable and usable in practice
- Convex developers using Claude Code or Cursor agents
- Teams building with Convex who want to help improve the platform
- Developers willing to share session transcripts for system improvement
- Users comfortable with opt-in telemetry for framework enhancement
convex-improve-convex-plugin FAQ
No. The helper script redacts keys and tokens before upload. Only the session transcript structure and interaction patterns are analyzed, not your application data.
Only once. The helper remembers your choice (Always / Just this once / Never) and respects it for future sessions.
The helper will print REVIEW_NO_TRANSCRIPT. This means no Claude/Codex .jsonl session file was detected in the current directory.
No. This is a system-improvement loop for Convex itself—identifying gaps in docs, skills, and tooling, not evaluating your coding.
The findings are generated server-side after submission. The helper prints status markers so you can track progress and see the summary when ready.
Full instructions (SKILL.md)
Source of truth, from get-convex/agent-skills.
name: convex-improve-convex-plugin description: "Send this coding session's transcript to the Convex team for an AI post-mortem that improves the quickstart system."
<!-- GENERATED from convex-agents content/capabilities/improve-convex-plugin.json — do not edit by hand. -->improve-convex-plugin
Sends the current coding session transcript to the anteater POST /review endpoint for an AI post-mortem. The review returns structured findings (ambiguous instructions, agent-stuck patterns, tooling failures, wins) targeted at the runbook, bootstrap script, skills, and components — not end-user data. Sharing is opt-in: the anteater-served helper asks once (Always / Just this once / Never) and remembers the choice.
Workflow
- Run the anteater-served helper:
curl -fsSL "<anteater>/send-transcript" | bash -s -- --idea "<one-line app idea from this session>". - If it prints CONSENT_REQUIRED (exit 4), the user has not chosen yet — ask them to share Always, Just this once, or Never, then re-run appending --consent always|once|never. Do not send until they answer.
- Watch for output markers: REVIEW_SOURCE (transcript found), REVIEW_SUBMITTED id=... (accepted), REVIEW_DONE status=done (findings ready).
- Summarize the highest-severity findings for the user: title → target → suggestedFix, then wins. Keep the summary about the system, not the user's data.
Rules
- Never send a transcript until the user has explicitly chosen to share (the helper prints CONSENT_REQUIRED and exits until they do).
- REVIEW_NO_TRANSCRIPT means no Claude/Codex .jsonl was found — tell the user.
- Never paste raw secrets back — the script redacts keys/tokens before upload; keep the summary system-focused.
- This is a system-improvement loop, not end-user feature feedback.
Related skills
More from get-convex/agent-skills and the wider catalog.
convex-insights
Query Convex app logs and health in natural language with evidence-backed answers and dashboard links.
convex-launch-readiness
Unified backend readiness score with prioritized fix plan—run all Convex audits, dedupe findings, and dispatch fixes.
convex-migrate
Safely migrate schema and backfill data on live Convex apps without downtime.
convex-migrate-rehearse
Rehearse schema changes on a preview deployment before promoting to production with snapshot rollback.
convex-migration-helper
Plan and execute safe Convex schema migrations with widen-migrate-narrow pattern and batched data backfills.
convex-monitor
Watch for the next error or request in a Convex app and react to it in real-time.