uditgoenka/autoresearch v1.0.2 on GitHub

What's New

/autoresearch:plan — Goal → Configuration Wizard

The hardest part of autoresearch isn't the loop — it's defining Scope, Metric, and Verify correctly. Get these wrong and the loop wastes iterations. Get them right and it's unstoppable.

/autoresearch:plan is an interactive wizard that converts your plain-language goal into a validated, ready-to-execute configuration.

/autoresearch:plan
Goal: Make the API respond faster

How It Works

The wizard walks you through 5-6 interactive questions to build a complete config:

Step	What Happens
1. Capture Goal	Ask what you want to improve (or accept inline text)
2. Analyze Context	Scan codebase for tooling, test runners, build scripts
3. Define Scope	Suggest file globs, validate they resolve to real files
4. Define Metric	Suggest mechanical metrics, reject subjective ones
5. Define & Validate Verify	Construct shell command, dry-run it, confirm it works
6. Confirm & Launch	Present config, option to launch immediately or copy

Critical Validation Gates

The wizard refuses to proceed until:

Gate	Requirement
Scope	Glob resolves to ≥1 file
Metric	Outputs a parseable number (not "PASS"/"FAIL"/subjective)
Verify	Dry-run succeeds (exit code 0 + extractable metric)

This means when you launch, you know the loop will work. No wasted iterations debugging a broken verify command.

Example Session

> /autoresearch:plan Reduce bundle size

[Scope]    Which files? → src/**/*.tsx, src/**/*.ts (127 files)
[Metric]   Bundle size in KB (lower is better)
[Verify]   npm run build 2>&1 | grep "First Load JS" | awk '{print $4}'
[Dry run]  ✓ Exit 0 — Baseline: 287KB

Ready-to-use:

  /autoresearch
  Goal: Reduce bundle size below 200KB
  Scope: src/**/*.tsx, src/**/*.ts
  Metric: bundle size in KB (lower is better)
  Verify: npm run build 2>&1 | grep "First Load JS" | awk '{print $4}'

Launch now? → [Unlimited] [Bounded] [Copy only]

Metric Suggestion Database

Includes domain-specific metric suggestions:

Domain	Metrics
Code quality	Coverage %, `any` count, lint errors, build errors
Performance	Bundle size, response time, Lighthouse score, build time
Content	Readability, word count, SEO score
Refactoring	LOC reduction, cyclomatic complexity, pattern elimination

6 Detailed Usage Examples

The README now includes full wizard flow examples for:

Test Coverage (Jest/Node.js) — npx jest --coverage
Lighthouse Performance (Next.js) — Lighthouse JSON output
TypeScript Strictness — grep-based any count
Python ML Training — val_bpb from training runs
Content SEO Scoring — custom scoring script
Docker Image Size — docker images size extraction

Each example shows the complete flow: goal → scope → metric → verify → dry-run → ready-to-paste config.

Launch Options

After validation, users choose:

Launch now — unlimited — start /autoresearch immediately
Launch now — bounded — specify N iterations via /loop N /autoresearch
Copy config only — get the ready-to-paste command block

Design Decisions

AskUserQuestion over free-text — structured choices reduce friction and prevent bad configs
Mandatory dry-run — the #1 failure mode is a broken verify command; dry-run catches this before loop starts
Metric suggestion DB — users shouldn't need to know what command to run; wizard infers from detected tooling
Max 5-6 questions — fast enough to not feel bureaucratic, thorough enough to catch misconfigurations
Inspired by the /plan --hard validate workflow pattern (question framework, validation gates, AskUserQuestion)

Files Changed

File	Change
`skills/autoresearch/SKILL.md`	v1.0.2: subcommands table, `/autoresearch:plan` docs, updated activation triggers
`skills/autoresearch/references/plan-workflow.md`	NEW — Full wizard protocol: 7 phases, validation gates, metric suggestion DB, error recovery
`README.md`	"Plan Your Run" section with 6 domain examples, new FAQ entry, version badge, updated repo structure

Installation

# Clone and copy to your project
git clone https://github.com/uditgoenka/autoresearch.git
cp -r autoresearch/skills/autoresearch .claude/skills/autoresearch

# Or copy to global skills
cp -r autoresearch/skills/autoresearch ~/.claude/skills/autoresearch

Pull Request: #2
Full Changelog: v1.0.1...v1.0.2

uditgoenka/autoresearch v1.0.2 v1.0.2 — Plan Your Run with /autoresearch:plan on GitHub