Rails Context Engineering

Load minimum context before any code, spec, or PRD in an existing Rails codebase. A fifteen-second read of db/schema.rb, config/routes.rb, and one neighbor saves a full retry.

HARD-GATE

DO NOT propose code, specs, PRDs, or task lists until the Context Summary is posted.
DO NOT silently resolve ambiguity — if requirements conflict or specs and code disagree, post a Confusion Block first.
DO NOT load the entire repo — use targeted reads (schema, routes, one neighbor of each kind).
ALWAYS cite the files you read (path:line where possible) so the user can verify.
ALWAYS re-check context when the user's request changes scope mid-conversation.

Process

1. Scope the change: In one sentence, name the Rails layer touched (controller, model, service, job, engine, view/Turbo, migration, API, GraphQL).
2. Load baseline Rails context: Read at minimum:
   • db/schema.rb — tables and columns involved (grep by table name)
   • config/routes.rb — routes that border the change
   • Gemfile.lock — confirm Rails version + domain gems (sidekiq, pundit, rspec, rails-i18n, graphql, etc.)
3. Load one neighbor of each kind: For each Rails layer touched, open the nearest sibling that already solves a similar problem — a comparable controller, service, spec, factory. Use grep to find: grep -r "class.*Controller" app/controllers, grep -r "class.*Service" app/services, etc.
4. Detect drift: If there is an existing spec for the area, compare what it asserts vs what the code currently does. Drift is a red flag — document it.
5. Post the Context Summary: Before any proposal, output the template below (see Output Style).
6. Handle ambiguity: If steps 2–4 surface a conflict (two patterns used, specs contradict code, missing requirement, unclear boundary), produce a Confusion Block:

### Confusion Block
- Conflict: <what conflicts — e.g., spec asserts X but code does Y>
- Options: <list the options with their tradeoffs>
- Recommendation: <state which option and why, or ask user to choose>

Do not pick silently. 7. Hand off: With context loaded, proceed to the next skill (create-prd, generate-tasks, rails-tdd-slices, rails-stack-conventions, etc.). The Context Summary travels with the task.

Output Style

Every invocation of this skill MUST produce a Context Summary in this exact shape, before any code or spec is proposed:

### Context Summary
- Scope: <one line — Rails layer and nearest class/file area>
- Rails version: <from Gemfile.lock>
- Relevant tables: <table names from db/schema.rb, columns that matter>
- Relevant routes: <resource/member routes from config/routes.rb>
- Nearest pattern: <path:line — one existing file that solves a similar problem>
- Nearest spec: <path:line — existing spec for this area (or NONE)>
- Engine boundary: <engine name or N/A>
- Gotchas: <domain gem quirks, enum mappings, polymorphic edges, counter caches, soft-delete, single-table inheritance — only list if present>
- Confusion: <NONE, or a one-line pointer to the Confusion Block below>

Additional MUSTs:

1. One neighbor per layer. Do not dump 5 similar files; pick the closest match and name it.
2. Facts only, no code. The summary lists facts about the codebase, not proposed implementation.
3. Hand-off line. End the reply with: Context loaded. Next: <skill-name> — <one-line reason>.

Pitfalls

Integration