skills.rest
dimagi-internal
dimagi-internalOFFICIAL
Explore all skills by this author →
GitHub

fork-run

Fork ACE opportunity runs at skill boundaries to preserve upstream artifacts.

1|2|Updated Apr 1, 2026
Search Tags:#run-artifacts#a-b-testing#ace#ace-web#fork-run#run-forking#phase-iteration
One-click install
npx skills add https://github.com/dimagi-internal/ace --skill fork-run
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: fork-run
Source: https://github.com/dimagi-internal/ace/tree/main/skills/fork-run
Command: npx skills add https://github.com/dimagi-internal/ace --skill fork-run

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

When iterating on ACE skills or recipes, re-running entire upstream phases to test a single downstream change wastes time and overwrites valuable prior run artifacts. This Skill eliminates that friction by letting you branch an existing run at a specific skill boundary, preserving all upstream work while providing a fresh run for downstream phases.

Core Features & Use Cases

  • A/B test skill or recipe changes: Fork a run, apply your modification to the forked run, and compare results against the original source run without altering its artifacts.
  • Iterate on a single phase: Re-run only the phase you are actively developing (e.g., Phase 6) without re-executing all earlier validated phases.
  • Head-to-head skill version comparison: Fork the same source run twice with different skill versions to directly compare their outputs and verdicts.

Quick Start

Use the fork-run skill to fork the turmeric opportunity's 20260513-2243 run at the app-test-cases skill boundary to re-run Phase 6 while preserving all upstream artifacts.

Frequently Asked Questions about fork-run

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

FAQPage Schema
How do I re-run a single downstream phase in ACE without overwriting prior run artifacts?▼

You can fork an existing ACE opportunity run at a specified skill boundary to isolate downstream phase re-runs. This preserves all upstream artifacts while providing a fresh run for the modified downstream phases.

What is run forking and when should I use it for A/B testing skill changes?▼

Run forking branches an existing run at a specific skill boundary to isolate downstream phases. Use it for A/B testing skill or recipe modifications and comparing skill versions without altering the original source run artifacts.

Do I need a personal access token to fork ACE opportunity runs?▼

Yes, forking ACE runs requires authenticated access to the ace-web fork endpoint via a personal access token. You also need a valid ACE opportunity slug, source run ID, fork skill name, and execution mode.

Can I compare two skill versions head-to-head using the same source run?▼

Yes, you can fork the same source run twice with different skill versions to directly compare their outputs and verdicts. This head-to-head comparison leaves the original source run artifacts completely unaltered.

Why does re-running an entire ACE workflow overwrite my validated upstream phases?▼

Re-running an entire workflow executes all phases sequentially and overwrites prior artifacts. Forking the run at the specific skill boundary you are developing preserves validated upstream phases and prevents overwrites.

Related Skills

Explore More Software Engineering Skills→
dimagi-internaldimagi-internal
1

sweep-opp-runs

Prune outdated per-opportunity run folders in ACE Google Drive.

Official
Advanced
matantsachmatantsach
1

run-evals

Run existing skill evaluations via npx snapeval eval and parse benchmark results.

Community
Intermediate
xdgxdg
20

isolated

Execute tasks in a forked context without affecting the main conversation.

Community
Basic
dimagi-internaldimagi-internal
1

cycle-grade-eval

Re-grade closed ACE cycle scorecards against raw run evidence.

Official
Advanced