Permissive start + progressive ratchet: CI-safe lint, presets, and ratchet DX

[mojoatomic]

Summary

Introduce a first-class "permissive start" mode and progressive ratcheting workflow to avoid CI breakage and improve onboarding.

Motivation

Current strict-by-default approach overwhelms teams (hundreds of violations) and breaks CI. We want:

• CI green on day 1
• Fix one category at a time
• Lock progress with ratchet
• Repeat

Scope

1. Permissive start (generator + preset)

• Init flag: --permissive / --permissive-start
• Generate flat config with measure-only rules (warn, not error)
• Include Jest/browser overrides to remove no-undef noise
• Export configs['permissive-start'] from the plugin for flat-config users

2. CI-safe lint + ratchet gate

• package.json scripts (generated):
  • lint: eslint .
  • lint:ci: eslint . || true
  • ratchet: eslint-plugin-ai-code-snifftest ratchet
• GitHub Actions snippet in docs: run lint:ci, then ratchet (only ratchet can fail)

3. Ratchet DX improvements

• Better fail message with: rule, baseline, current, first N new occurrences (file:line), and "how to fix" tip
• Add --unlock=<rule> path documented in message
• Script alias: ratchet:explain to print baseline JSON

4. Issue creation hygiene (optional, separate PR if large)

• create-issues --phase=1|2|3 to avoid dumping all issues at once
• Priority labels in titles: [P0-Quick-Win], [P1-High-Impact], etc.
• START HERE issue or TECH-DEBT.md scaffold with weekly trend

Acceptance Criteria

• init --permissive generates permissive ESLint config (warn-only) + Jest/browser overrides
• Plugin exports configs['permissive-start']
• Generator adds lint:ci + ratchet scripts and docs snippet
• Ratchet prints actionable failure messaging (with example and unlock hint)

Non-Goals

• Changing default behavior for existing strict users
• Enforcing env-based severity switching (prefer ratchet gating)

Notes

This enables: CI stays green, teams fix one rule at a time, and ratchet prevents backsliding.