Whistler Snow Report Automation Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Add a Ruby CLI that turns /tmp images plus a short summary into a finished Whistler Blackcomb snow report with low-token AI generation and validation.

Architecture: Keep deterministic work local in Ruby: discover the report date, resize images, choose the previous featured post, and write the markdown file. Use codex exec with gpt-5.1-codex-mini only for body generation and a second low-cost pass for style/fact review, both driven by a short style capsule and a small reference excerpt. Keep the existing routine skill as a thin pointer to the script so future runs are short.

Tech Stack: Ruby, existing image.sh, existing bin/snow-report-lint, existing bin/snow-report-review, Codex CLI.

Task 1: Add a Ruby CLI for report creation

Files:

• Create: bin/snow-report-create

• Step 1: Write the failing test

1
2
# No automated test file exists yet; this CLI will be verified by running it
# against a temporary set of /tmp images and checking the generated post.

• Step 2: Run the command with no images

Run: bin/snow-report-create --summary "rate 4, partially sunny & cloudy, not busy"

Expected: fail fast with a clear message explaining that /tmp images are required.

• Step 3: Write minimal implementation

1
#!/usr/bin/env ruby

• Step 4: Run the command with sample inputs

Run: bin/snow-report-create --date 2026-03-31 --summary "rate 4, partially sunny & cloudy, not busy"

Expected: create _posts/2026-03-31-2026-03-31-whistler-blackcomb-snow-report.md, resize /tmp images into assets/images/, and set the previous featured report to featured: false.

• Step 5: Commit

1
2
git add bin/snow-report-create
git commit -m "Add snow report creation CLI"

Task 2: Tighten the routine skill to point at the script

Files:

• Modify: .agent/skills/routine/SKILL.md

• Step 1: Replace the long manual checklist with a short usage note

1
Use `bin/snow-report-create --summary "..."` after placing images in `/tmp`.

• Step 2: Keep the key invariants

1
The script handles image conversion, featured flipping, linting, review, and commit.

• Step 3: Commit

1
2
git add .agent/skills/routine/SKILL.md
git commit -m "Simplify snow report routine"

Task 3: Verify the low-token AI path

Files:

• Modify: bin/snow-report-create

• Step 1: Run the CLI end-to-end with a small image set

Run: bin/snow-report-create --date 2026-03-31 --summary "rate 4, partially sunny & cloudy, not busy"

Expected: codex exec runs twice with gpt-5.1-codex-mini, first for body generation and then for a short review pass.

• Step 2: Run lint and review

Run: bin/snow-report-lint --auto-correct && bin/snow-report-review

Expected: both commands report ok.

• Step 3: Commit

1
2
git add bin/snow-report-create _posts/2026-03-31-2026-03-31-whistler-blackcomb-snow-report.md assets/images/2026-03-31-feature.jpg .agent/skills/routine/SKILL.md
git commit -m "Automate Whistler snow report creation"