Initialize this-little-wiggy configuration
Interactive wizard that configures ralph-loop completion criteria for your project.
/plugin marketplace add severity1/this-little-wiggy/plugin install severity1-this-little-wiggy@severity1/this-little-wiggy[--force]Interactive wizard to configure ralph-loop for this project.
REQUIRES user confirmation via AskUserQuestion before writing config.
if [ -f ".claude/this-little-wiggy/config.yml" ]; then
echo "CONFIG_EXISTS"
cat .claude/this-little-wiggy/config.yml
else
echo "NO_CONFIG"
fi
If config exists and --force was NOT provided in $ARGUMENTS:
First, explain to the user:
"this-little-wiggy wraps your prompts with completion criteria so ralph-loop knows when a task is truly done. We'll auto-discover tests, linters, and quality checks from your project to build these criteria."
Use AskUserQuestion (1 question):
Ask: "Does this project have tests or quality checks we can discover?"
Options:
If user selects "No, I need to set those up first":
/this-little-wiggy:init again when ready."For other responses, tell them:
/this-little-wiggy:init --force."Use the Task tool with subagent_type: Explore (thoroughness: "very thorough") to discover what "done" looks like for this project.
Prompt the Explore agent:
Thoroughly explore this codebase to find ALL quality checks and completion criteria. Be exhaustive.
Look for:
- All Makefile targets (not just build/test/lint)
- All test types: unit, integration, golden/snapshot, e2e
- Static analysis: vet, fmt, gocyclo, complexity checks
- Linting configs and all linter rules
- CI/CD pipelines - what checks run there?
- .claude/ directory: skills, commands, slash commands (like /go-review)
- CLAUDE.md and any documentation mentioning required checks
- Any pre-commit hooks or git hooks
- Code review requirements
Return a comprehensive list of everything that should pass before work is considered done.
Build a list of completion criteria from the agent's findings.
If no criteria discovered:
STOP - Use AskUserQuestion before proceeding.
First AskUserQuestion call (2 questions):
Max iterations (multiSelect: false)
Core checks (multiSelect: true) - build, test, lint, coverage
Second AskUserQuestion call (up to 3 questions): 3. Static analysis (multiSelect: true) - vet, fmt, gocyclo, etc.
Advanced tests (multiSelect: true) - integration, golden, e2e, etc.
Other (multiSelect: true) - review commands, docs, etc.
Only include questions where criteria were discovered.
WAIT for all responses before continuing.
mkdir -p .claude/this-little-wiggy
Write .claude/this-little-wiggy/config.yml with user's selections:
defaultMaxIterations: 10
ralphWrapper: |
/ralph-loop:ralph-loop --max-iterations 10 --completion-promise "COMPLETE" "{task}
When complete:
- Tests pass
- Lint passes
- Build succeeds
Output <promise>COMPLETE</promise> when done."
IMPORTANT:
.claude/this-little-wiggy/config.yml{task} placeholder is substituted at runtimemake test")Output summary:
.claude/this-little-wiggy/config.ymlNext steps:
/this-little-wiggy:init --force to regenerate