convex-self-heal

Triages production errors and prepares certified fix PRs for human review.

9.4k|1.5k|Updated Jan 3, 2026
One-click install
npx skills add https://github.com/openclaw/clawhub --skill convex-self-heal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: convex-self-heal
Source: https://github.com/openclaw/clawhub/tree/main/.agents/skills/convex-self-heal
Command: npx skills add https://github.com/openclaw/clawhub --skill convex-self-heal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Production errors in Convex deployments typically require manual investigation, root-cause analysis, and unverified fixes that engineers must debug under pressure. This Skill automates the loop from error capture to a certified fix PR while keeping the human in control of the merge.

Core Features & Use Cases

  • Error Triage and Root-Cause Analysis: Classifies production errors as transient, config, or code/schema defects, then runs audit passes (convex-insights, convex-advisor, convex-reviewer) to produce evidence-backed findings.
  • Certified Fix Preparation: Repairs defects on a branch and certifies them with tsc checks, schema migration rehearsal on a prod-snapshot preview, and reproduce-then-confirm-gone testing before proposing anything.
  • Human-Gated Promotion: Opens a PR with full certification evidence and never auto-merges; after merge and deploy, it verifies the error signature stops recurring in production.
  • Use Case: A recurring production error appears in your Convex deployment's sentinel table. The Skill triages it, root-causes a missing ownership check, certifies the fix against real data invariants, and opens a PR you simply review and merge.

Quick Start

Ask the agent to investigate the latest recurring production error from the sentinel table and prepare a certified fix PR for review.

Frequently Asked Questions about convex-self-heal

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically fix production errors in a Convex backend?

Use a gated self-healing loop that captures errors via a sentinel table, triages them by type, root-causes code defects with audit passes, and prepares a certified fix PR. The human always keeps the merge button; the system never auto-merges or auto-deploys.

How does the self-healing loop certify a fix before proposing it?

Certification requires a clean tsc --noEmit run, schema migration rehearsal on a preview seeded with a prod snapshot when data changes are involved, and a reproduce-then-confirm-gone test replaying the triggering input. Fixes failing any applicable check are reported as uncertified, never proposed.

Does this workflow auto-merge or auto-deploy fixes to production?

No. The loop only prepares and certifies fixes on a branch, then opens a PR for human review. Promotion happens only through explicit deploy consent after the human merges, matching the industry boundary against unattended prod auto-merge.

What types of production errors can be auto-prepared versus deferred?

Pre-approved safe classes include validator fixes, missing-index additions, ownership-check additions, and non-destructive backfills. Destructive changes, security-sensitive modifications beyond added checks, and ambiguous cases are always deferred to explicit human direction.

How is it confirmed that a fixed error stopped recurring in production?

After merge and deploy, the loop re-checks the sentinel table and failure logs to confirm the error signature no longer recurs. It does not use insights for this, since insights only tracks OCC and read-limit performance events, not arbitrary error signatures.