skills.rest
PrimeIntellect-ai
PrimeIntellect-aiOFFICIAL
Explore all skills by this author →
Website𝕏@primeintellectGitHub

cleanup-tasks

Diagnose and repair broken general-agent task definitions failing validation.

113|35|Updated Nov 29, 2025
Search Tags:#automation#debugging#task validation#triage#repository maintenance#json replay#verify() scoring
One-click install
npx skills add https://github.com/PrimeIntellect-ai/research-environments --skill cleanup-tasks
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: cleanup-tasks
Source: https://github.com/PrimeIntellect-ai/research-environments/tree/main/environments/general_agent/skills/cleanup-tasks
Command: npx skills add https://github.com/PrimeIntellect-ai/research-environments --skill cleanup-tasks

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill helps you diagnose and fix broken tasks in environments/general_agent/tasks when validation fails because gold.json can’t be replayed on db.json or when verify() rejects the replayed state.

Core Features & Use Cases

  • Classify failures by running general-agent validate and grouping failures to determine the most common root causes.
  • Diagnose failing tasks using a helper script that reports the failure class, the first actionable error message, and a targeted hint.
  • Apply repair heuristics or delete unrecoverable tasks to restore a clean validation state after synthesis, refactors, or periodic sweeps.

Quick Start

Diagnose failing tasks by running uv run general-agent validate --fail-only and then passing the resulting task names to uv run python environments/general_agent/skills/cleanup-tasks/diagnose.py.

Frequently Asked Questions about cleanup-tasks

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

FAQPage Schema
How do I fix broken task definitions when general-agent validate reports failures?▼

To fix broken task definitions, run the validation command to isolate failures, then pass the failing task names to the diagnose script to identify failure classes and apply targeted edits or delete unrecoverable tasks.

Why does gold.json replay fail to match my db.json state during task validation?▼

Gold.json replay fails to match db.json state due to mismatches occurring after synthesis or refactors, resulting in failure classes like gold replay issues, gold no-change, or verify() scores falling below 1.0.

How do I diagnose verify() score failures below 1.0 in task maintenance?▼

Diagnose verify() score failures by running the validation command with the fail-only flag, then passing failing task names to the diagnose script to receive actionable error messages and targeted hints for repair.

What is the best way to triage broken tasks after a repository maintenance sweep?▼

Triage broken tasks after a maintenance sweep by snapshotting failing definitions, classifying their root causes using the diagnose script, and applying repair heuristics or deleting unrecoverable tasks to restore clean validation.

When should I delete unrecoverable task definitions instead of attempting a shallow fix?▼

Delete unrecoverable task definitions instead of applying shallow fixes when the diagnose script determines the failure class is unrecoverable, ensuring re-validation passes cleanly after rescoping or removing broken tasks.

Related Skills

Explore More Software Engineering Skills→
PrimeIntellect-aiPrimeIntellect-ai
113

synthesize-task

Generate tiered general-agent task families with empirical pass-rate gating.

Official
Advanced
gaiastudio-aigaiastudio-ai
2

gaia-fix-story

Fix open validation findings in story files and re-validate until ready-for-dev.

Official
Intermediate
nicsuzornicsuzor
3

sleep

Periodically consolidates AI agent activity through session backfill, episode replay, index refresh, staleness sweeps, and brain sync via cron or manual invocation.

Community
Intermediate
nimrodfishernimrodfisher
351

metric-reconciliation

Compare cross-source metrics and generate a reconciliation report.

Community
Intermediate