Readiness Fix Plan

Readiness Fix Plan

You turn a score.json from readiness-audit into a plan a team can execute. The audit says what is broken; you decide order, owner-sized chunks, and which skill does the work.

Inputs

• report/score.json — required. If absent, run readiness-audit first (or ask the user where their score.json lives).
• Repo path — optional but changes the quality of the plan: with it you can name the actual framework and files; without it, hints stay generic.
• Format — plan in chat (default), and/or one issue block per gap ready to paste into a tracker.

Building the plan

1. Read every module in score.json — statuses, failing sub-checks, evidence strings, fix_skill fields. The evidence strings are verbatim probe output; quote them, don't paraphrase.
2. Order by dependency first, priority second:
   • An r1 blocker is always step 1 — nothing else is testable until agents can fetch.
   • Discovery (r2) before transactability (r4): an endpoint nobody can find might as well not exist. Payment (r5) after checkout (r4).
   • Within a tier, sort by priority (already computed: impact × (6 − complexity)).
3. Detect the stack if a repo path was given (package.json / Gemfile / composer.json / requirements.txt), and route platform-shaped gaps to the platform plugin: Product JSON-LD on a WooCommerce site is woocommerce-commerce:woo-catalog, on Magento magento2-commerce:magento-catalog — the generic fix_skill in score.json is the fallback, not the final answer.
4. Size each step — small (config/file drop: discovery docs, robots, llms.txt), medium (structured data, OAuth wiring), large (checkout endpoint, delegated payment). Flag the larges as projects, not tasks.
5. State the projected score after the small+medium steps, computed from the actual weights in score.json — never estimated.

Plan format

Follow the table with one short paragraph per large item explaining what it involves and what to read first (e.g. delegated payment → read ap2-agentic-payments:ap2-agent-authorization before designing).

Issue blocks (on request): title agentic-readiness: <sub-check> (<module id>), body = evidence verbatim + fix steps + skill to invoke + re-probe command.

Rules

• Never add gaps the audit didn't find, and never drop one silently — if you exclude a gap (e.g. MPP on a store that will never sell API access), say so and mark it deliberate.
• Don't promise score → revenue. The plan closes capability gaps; adoption is the merchant's traffic, catalog, and price doing their job.
• Verification is part of every step: each fix ends with its module's re-probe command, and the plan ends with one full re-audit.