nuttaruj

Use when executing an approved plan or a clear single-file edit — TDD for risky paths, surgical edits, bounded delegation, worktrees only when real filesystem isolation is needed. Phase = Build.

Use when executing an approved plan or a clear single-file edit — TDD for risky paths, surgical edits, bounded delegation, worktrees only when real filesystem isolation is needed. Phase = Build.

Use when executing an approved plan or a clear single-file edit — TDD for risky paths, surgical edits, bounded delegation, worktrees only when real filesystem isolation is needed. Phase = Build.

Use at the end of a development branch — pre-merge gate, CI lane discipline, 4-option finish menu (merge, PR, keep open, discard), release checklist for production launches. Phase = Ship.

Use at the start of every request to route work into Rolepod's workflow spine before planning, editing, delegating, verifying, reviewing, or shipping. Determines phase, required skills, skip rules, and evidence needed.

Use when turning an approved spec or a small clear goal into an executable implementation plan — ordered tasks, file list, test plan, agent routing, and parallel contracts if more than one agent will edit code. Phase = Plan.

Use when turning a fuzzy goal, half-stated feature, or vague request into a sharp implementation spec. Discovery dialogue first, then design, then user approval, then a compact contract. Phase = Define.

Use at the end of a development branch — pre-merge gate, CI lane discipline, 4-option finish menu (merge, PR, keep open, discard), release checklist for production launches. Phase = Ship.

Create, read, update, search posts/pages/users/options/taxonomies via the core WP REST API on a connected target. Phase = Build.

Read or write configuration owned by SEO / i18n / e-commerce / custom-field / forms plugins — Yoast, RankMath, WPML, WooCommerce, ACF, Gravity / CF7 / WPForms — on a connected target. Phase = Build.

Recover a WordPress site that is white-screening / 500-ing / fatal after an edit — guardian mu-plugin, safe-mode, disable/restore the culprit — plus safe core/plugin/theme updates and guarded .htaccess edits. Phase = Recover.

Multi-probe diagnostic + security audit of a connected target — plugin conflicts, slow queries, large autoload options, broken images, PHP errors, outdated core/plugins/themes, weak admins. Phase = Debug.

Plan, dry-run, apply, back up, and clone migrations between two WP targets — dev → staging → prod, with rollback. Phase = Ship.

Query the AI Change Ledger on a connected target — the writes the MCP recorded through the companion (RestTarget only), categorized; toggle on/off; panic-revert recent changes. Phase = Recovery.

Read or modify page-builder layouts on a connected target — Elementor, Divi, Oxygen, Bricks widget trees. Phase = Build.

Open a WordPress Target on any kind (local / REST + App Password / SSH / Docker) using credentials already in the vault. Phase = Define.

Run an axe-core accessibility audit on a page against WCAG-A / WCAG-AA / WCAG-AAA and return issues grouped by severity with WCAG references and fix links. v0.2 web only; scope='page' only.

Audit a page's weight against a declared byte budget. Loads the URL, records a HAR, classifies entries by asset category (js/css/image/font), tags third-party requests, and compares totals to budget. Returns violations + graduated pass/warn/fail status. Driven by Playwright, observed entirely in-browser.

Audit on-page SEO by inspecting the rendered DOM. Checks title, meta description, h1 structure, html lang, viewport, canonical, robots, OpenGraph + Twitter Card tags, JSON-LD validity, hreflang, and favicon. Returns findings grouped by severity (critical / high / medium / low).

Drive a flow and fail if any console error or failed network request occurs. Thin wrapper over verify_ui_flow with strict error-only assertions. Use to gate merges on "no regressions during this flow".

Measure Core Web Vitals (LCP, INP, CLS) on a live page using PerformanceObserver injection. Returns metrics + thresholds verdict per web.dev good/needs-improvement/poor bands. Chromium-only — Firefox and WebKit ship partial PerformanceObserver coverage.

Generate a runnable e2e test file (playwright-test, vitest+playwright, or pytest+selenium) from a scenario description plus an optional replay bundle from a prior /verify-ui run.

Drive a real browser session through steps and assert expected outcomes — including console errors, network failures, and visual state. Save evidence under ./.rolepod-uiproof/artifacts/. Web only.

Capture a screenshot of the current UI and compare against a stored baseline under ./.rolepod-uiproof/baselines/. First capture for a baseline_id seeds the baseline; subsequent capture diffs it.

Run EXPLAIN [ANALYZE] on a query and return the parsed plan with flags for sequential scans and probable missing indexes — performance evidence, not a guess. Read-only. Use when a query is slow or before shipping a query-heavy path.

Read a live Postgres schema — tables, columns, types, primary keys, indexes, and the foreign-key graph — into a normalized snapshot for planning and debugging. Read-only. Use before changing data access, writing a migration, or reasoning about how tables relate.

Detect schema drift between SQLAlchemy models and the live Postgres — missing/extra columns, nullability mismatches, type mismatches. Reflects the models via a Python sidecar and diffs in TS. Read-only. Use as a pre-ship gate or when "the code says one thing, the DB says another".

Assert live database state as verify-phase evidence. Run a read-only query and check it against an expected exists / row_count / value, returning a structured PASS/FAIL — not raw rows. Use to prove a row exists, a count matches, or a value landed after a change.

Mutate data behind a transaction guard — preview a statement inside a held transaction (no commit), inspect the impact, then confirm to COMMIT or rollback to discard. The only dblab skill that writes. Use for any INSERT/UPDATE/DELETE that must be safe and reviewable.