verify-fix-loop

Coordinate bounded fix-verify iterations with EvidenceBundle updates and escalation policies.

15|4|Updated Mar 26, 2026
One-click install
npx skills add https://github.com/choam2426/geas --skill verify-fix-loop
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: verify-fix-loop
Source: https://github.com/choam2426/geas/tree/main/plugin/skills/verify-fix-loop
Command: npx skills add https://github.com/choam2426/geas --skill verify-fix-loop

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

After the evidence gate fails, this loop ensures bugs are actually fixed — not just "code is written." It coordinates bounded retry iterations to verify genuine correctness before proceeding to the Closure Packet.

Core Features & Use Cases

  • Read TaskContract for retry_budget and escalation_policy to determine iteration limits.
  • Produce per-iteration EvidenceBundle updates (fix-{N}) and update the current run state.
  • Escalate via DecisionRecord when retry_budget is exhausted and log escalation events.
  • Spawn implementers with worktree isolation to perform fixes in an isolated context.

Quick Start

Trigger the verify-fix-loop after a gate failure to iteratively apply fixes up to the configured retry_budget.

Frequently Asked Questions about verify-fix-loop

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

FAQPage Schema
How do I automate an iterative fix and verify loop for failed tasks?▼

To automate a fix and verify loop, coordinate bounded retry iterations that read the TaskContract for retry_budget, update EvidenceBundle per iteration, and invoke fixers until passing verification or exhausting the budget.

What happens when a retry budget is exhausted during automated bug fixing?▼

When the retry budget is exhausted during automated bug fixing, the loop executes the escalation_policy defined in the TaskContract, records a DecisionRecord, and optionally pivots to higher authority for resolution.

How does worktree isolation improve quality assurance during code fixes?▼

Worktree isolation improves quality assurance by spawning implementers in an isolated context to perform fixes, preventing interference with the main branch while generating evidence for the verification gate.

Do I need a TaskContract to run automated fix verification loops?▼

A TaskContract is required to run fix verification loops because it supplies the retry_budget and escalation_policy parameters that determine iteration limits and fallback escalation procedures.

Can I use an evidence gate to verify code correctness before closure?▼

Yes, you can use an evidence gate to verify code correctness before closure by returning per-iteration EvidenceBundle updates to the gate until a PASS verdict is reached or the retry budget is depleted.

Why does my fix loop keep failing instead of escalating after retries?▼

A fix loop fails to escalate when the TaskContract lacks a valid escalation_policy or the retry_budget is misconfigured. Upon budget exhaustion, the loop should record a DecisionRecord and execute the escalation policy.