Resume
Resume a paused autoresearch experiment branch, reload config, program, and results.tsv, then continue metric-driven iterations without losing history.
Overview
Resume is an agent skill most often used in Operate (also Build) that restores a paused autoresearch experiment branch and continues metric-driven iterations from saved results history.
Install
npx skills add https://github.com/alirezarezvani/claude-skills --skill resumeWhat is this skill?
- Lists autoresearch experiments via setup_experiment.py when no path is given
- Checkouts autoresearch/{domain}/{name} and loads config, program.md, and full results.tsv
- Summarizes target file, metric direction, experiment counts, and best result vs baseline
- Surfaces recent git log and pattern notes (e.g. caching vs algorithm changes)
- Continues the /ar experiment workflow from the last recorded outcome
- Example state report cites 23 experiments (8 kept, 12 discarded, 3 crashed)
- Loads up to 20 recent commits via git log --oneline
Adoption & trust: 1.4k installs on skills.sh; 17.5k GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
You paused or ran out of context mid-experiment and cannot quickly reconstruct branch state, metrics, and which changes were kept or discarded.
Who is it for?
Solo builders using autoresearch branches with results.tsv who need a fast handoff back into iterative performance or behavior tuning.
Skip if: Greenfield projects without autoresearch folders, one-off edits with no experiment branch, or teams that do not use git-based experiment tracking.
When should I use this skill?
A paused or context-limited autoresearch experiment needs to continue with full history on autoresearch/{domain}/{name}.
What do I get? / Deliverables
You get a structured state report from config, program.md, results.tsv, and recent commits, then continue experimenting on the same branch with the same target and metric.
- Human-readable experiment state summary (metric, counts, best result, recent patterns)
- Checked-out experiment branch ready for the next iteration
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Canonical shelf is Operate because the skill’s job is continuing an in-flight improvement loop after pause or context limits—not starting research or shipping a release. Iterate subphase fits listing experiments, reporting keep/discard/crash stats, and picking up from the last KEEP row on the branch.
Where it fits
After a weekend away, checkout autoresearch/engineering/api-speed and report best p50_ms before the next caching tweak.
Resume tuning src/api/search.py with full keep/discard history instead of guessing which changes helped.
Re-open a latency experiment branch to push further wins after a partial deploy baseline.
How it compares
Workflow handoff for paused experiments—not a generic git stash pop or ad-hoc “where were we?” chat recap.
Common Questions / FAQ
Who is resume for?
Indie and solo developers running autoresearch or similar git-branch experiments who need the agent to reload history and continue iterating.
When should I use resume?
In Operate when you return after a break or context limit; in Build when you are still tuning a target file (e.g. API latency) and want the last KEEP/DISCARD trail before the next change.
Is resume safe to install?
It runs git checkout and reads project files; review the Security Audits panel on this page and confirm branch names before letting an agent switch branches.
SKILL.md
READMESKILL.md - Resume
# /ar:resume — Resume Experiment Resume a paused or context-limited experiment. Reads all history and continues where you left off. ## Usage ``` /ar:resume # List experiments, let user pick /ar:resume engineering/api-speed # Resume specific experiment ``` ## What It Does ### Step 1: List experiments if needed If no experiment specified: ```bash python {skill_path}/scripts/setup_experiment.py --list ``` Show status for each (active/paused/done based on results.tsv age). Let user pick. ### Step 2: Load full context ```bash # Checkout the experiment branch git checkout autoresearch/{domain}/{name} # Read config cat .autoresearch/{domain}/{name}/config.cfg # Read strategy cat .autoresearch/{domain}/{name}/program.md # Read full results history cat .autoresearch/{domain}/{name}/results.tsv # Read recent git log for the branch git log --oneline -20 ``` ### Step 3: Report current state Summarize for the user: ``` Resuming: engineering/api-speed Target: src/api/search.py Metric: p50_ms (lower is better) Experiments: 23 total — 8 kept, 12 discarded, 3 crashed Best: 185ms (-42% from baseline of 320ms) Last experiment: "added response caching" → KEEP (185ms) Recent patterns: - Caching changes: 3 kept, 1 discarded (consistently helpful) - Algorithm changes: 2 discarded, 1 crashed (high risk, low reward so far) - I/O optimization: 2 kept (promising direction) ``` ### Step 4: Ask next action ``` How would you like to continue? 1. Single iteration (/ar:run) — I'll make one change and evaluate 2. Start a loop (/ar:loop) — Autonomous with scheduled interval 3. Just show me the results — I'll review and decide ``` If the user picks loop, hand off to `/ar:loop` with the experiment pre-selected. If single, hand off to `/ar:run`.