Resume Paused Autoresearch Experiments
Resume a paused autoresearch experiment from its verified branch and results history without losing prior trials.
Before & after
Same prompt, same agent — the only difference is the skill.
Sure, I can help you get back up to speed on your feature. Let me check the current state of your branch, look at recent commits, and see what you were working on. I'll need to find the branch name and pull the latest changes. Then I can review what's been done, any test results, and what the next steps might be. Do you remember what the feature was called, and do you have any notes about where you left off or what was blocking you?
FEATURE RESUME SNAPSHOT BRANCH: feat/real-time-collab-editor LAST COMMIT: "wip: WebSocket transport layer" (2026-07-28 14:30) COMMITS SINCE: 0 (paused) PROGRESS STATUS: COMPLETE: ✓ Project initialized with EF Core DbContext ✓ User authentication middleware wired ✓ Document API endpoints (GET, POST, PUT) ✓ WebSocket server booted; test client connects IN PROGRESS: • WebSocket message routing (70% done) • Conflict resolution logic (pseudo-code, untested) BLOCKED: • Persistence layer needs schema review by lead dev • Merging main branch (3 new commits conflict) OPEN PRS: • #247 "Add WebSocket transport" • #251 "Schema updates for collab" TEST STATUS: • Unit tests: 12 passing, 3 skipped (conflict resolution edge cases) • Integration tests: pending WebSocket fixes NEXT STEPS: 1. Merge main (resolve conflicts in migrations) 2. Finish conflict resolution tests 3. Load-test with 5 concurrent editors 4. Unblock lead dev schema review ESTIMATED COMPLETION: 5 days if unblocked.
About this skill
“Pick up where we left off” gets risky when the latest commit, best result, and unfinished trial are different states. This skill reconstructs the experiment from its branch and complete results history, verifies the checkpoint, and continues without silently changing the benchmark or repeating completed work.
What it does
- Identifies the correct experiment using branch, objective, evaluation command, and results records.
- Reconciles completed trials with commits, measured scores, and acceptance decisions.
- Distinguishes the latest attempted change from the best accepted checkpoint.
- Recovers interrupted trials without duplicating active jobs or treating partial scores as final results.
- Checks for changes to evaluation inputs, dependencies, and measurement conditions before comparing scores.
- Continues within the existing scope and budget, then records a checkpoint that can be resumed again.
What you get
A concise recovery readout identifying the verified branch, accepted checkpoint, unresolved work, and next trial. When execution is available and authorized, you also get the resumed trial, its measured outcome, and an updated experiment record.
Best for
Developers using Claude Code to resume repository-based optimization experiments after interruptions, context loss, branch switches, or unfinished evaluation runs.
SKILL.md preview
---
name: resume-paused-autoresearch-experiments
description: Recover a paused autoresearch experiment by reconciling its branch, complete results history, accepted checkpoint, and interrupted work, then continue its existing evaluation protocol. Use when the user says "resume the experiment", "pick up the autoresearch run", "/ar:resume", "continue from the last good result", "my experiment stopped halfway through", "the results log doesn't match the branch", or "which paused experiment were we running?".
---
# Resume Paused Autoresearch Experiments
… (sign up to view the full skill)More development skills
View all Development skills →Performance Bottleneck Profiling: Node, Python, Go
Classifies a slowdown, gives exact profiler commands, reads the output, and proves the fix with same-conditions p95/p99 numbers.
Feature Scaffolding with Working Tests
Turn a short feature spec into code, runnable tests, and integration wiring that match your existing project.
Feature Flag Rollouts and Kill Switches
Design staged flag rollouts, verify rollback paths, and audit flag debt with explicit defaults, gates, and retirement plans.
Microcontroller Firmware and RTOS Engineering
Build and debug MCU firmware with bounded ISR/RTOS timing, safe DMA ownership, and measured power budgets.