dotfiles-claude

Deploy and verify the Bella Sláinte marketing site (BellaSlainteSite) on Vercel. Use when deploying bellaslainte.com, shipping site HTML changes, checking if the site is live, fixing blocked Vercel deploys, linking GitHub to Vercel for haberlah/david-6804, or running vercel CLI against project-iw363. Triggers: "deploy the website", "ship BellaSlainteSite", "push site to prod", "is bellaslainte.com live", "Vercel blocked", "redeploy marketing site", /vercel-bellaslainte-site.

This skill should be used whenever a task involves creating, changing, or granting access to any GCP resource — creating a project, adding or modifying an IAM binding or role grant, provisioning a secret, KMS key, or CMEK, enabling an API, changing an org policy, or touching networking, folders, or service accounts. It enforces that every such write happens through FAST (Fabric Automation and Scalability Toolkit / Cloud Foundation Fabric) YAML plus Terraform plus a reviewed PR against the foundations repo, never via ad-hoc `gcloud` commands or GCP Console clicks — including when the request is phrased as "quickly," "just for now," "as a one-off," or "to unblock." Also use this skill when asked to bring an already-existing (brownfield) GCP resource under Terraform management, or when reviewing whether a proposed GCP change is safe to apply directly. Read-only/diagnostic GCP requests (describing a project, listing IAM bindings, checking quota, viewing Console screens) do not need this skill.

Use whenever working on BellaAssist (Bella Assist) product, runtime, schema, API, AI skills, auth/RLS, CRM integration, ops, compliance registers, or architecture documentation — including phrases like "architecture pack", "arch-pack", "arch pack", "current-state docs", "how does X work in Bella", "endpoint inventory", "schema inventory", or "before coding BellaAssist". Routes agents to the BellaAssist architecture-pack for business logic maps, schemas, and generated inventories, and enforces the regenerate → validate → snapshot-sync loop. Also use when updating docs after a BellaAssist behaviour change, checking documentation drift, or deciding which regression tests to run.

Operate David's Jira Cloud (bellaslainte.atlassian.net) from the terminal with the official Atlassian CLI `acli`. Use whenever the task involves reading or writing a Jira board/project, creating/editing/transitioning/deleting work items, populating a Kanban board (e.g. from a Google Sheet), running JQL searches, or installing/authenticating acli. Captures the non-obvious gotchas: search defaults to 30 results, create-bulk caps at 50 and silently drops labels + ignores description in JSON, edit --labels appends, and the KAN board's columns are Idea/In Progress/Done.

How to authenticate and operate the AWS CLI for David's Bella Slainte AWS account (571573791861), and how to manage AWS API keys / IAM / IAM Identity Center. Use whenever the task involves running AWS CLI commands, logging in to AWS or refreshing AWS credentials, creating or rotating API keys / IAM access keys / Bedrock API keys, adding or auditing IAM Identity Center (SSO) users and permission sets, or any AWS console/CLI administration. Covers the SSO login gotchas (portal username is "David" not his email; the `aws configure sso` region prompt loops).

Review GitHub pull requests against a configurable AI review policy for Codex and Claude Code. Use before creating, triggering, checking, re-running, or summarizing PR reviews; when deciding between @codex review and @claude review once; when inspecting bot review results, GitHub checks, branch protection, CODEOWNERS, or merge readiness; and when verifying whether generic thumbs-up/no-findings review text actually means the review bot ran.

Commit, push, and optionally create a PR. Use after completing a task that changes files in a project repo or when the user asks to push, commit, or create a PR. Invoke GitHub PR review policy before any review trigger. Paid Claude Code Review requires explicit wording such as "Claude Code Review", "manual @claude review", or selecting the Claude review menu option; generic "review it" is not enough.

Authenticate or repair Google Workspace CLI (`gws`) OAuth with reusable scope presets instead of manually toggling scopes. Use when setting up GWS CLI, running `gws auth login`, fixing invalid_scope errors, avoiding `cloud-identity.devices`, refreshing invalid tokens such as invalid_rapt, or preparing a broad Google Workspace CLI OAuth login for engineering workflows.

Read, write, search, and manage David's Obsidian vault via the `obsidian` CLI installed at /opt/homebrew/bin/obsidian. Use when the user wants to save a note to Obsidian, read an existing note, search the vault, append to a daily note, manage vault content, or asks about Obsidian content. Vault path is `~/Documents/Obsidian Vault/`. The CLI is bundled with Obsidian.app and exposes 90+ commands covering files, folders, daily notes, tags, properties, search, bookmarks, templates, themes, plugins, and Obsidian Sync. CRITICAL: the CLI requires "Settings → Community plugins → Browse → install Obsidian CLI" to be enabled in Obsidian itself. Direct file write to the vault folder is the simpler fallback when scripted-vault automation is overkill.

Keep long Codex conversations alive: repair an oversized session that crashes the Codex desktop app with "RangeError: Invalid string length" or won't open a conversation. Use whenever Codex won't reload a session, the desktop app crashes or white-screens on launch or when opening a chat, or a session rollout file under ~/.codex/sessions has grown huge (commonly from embedded base64 screenshots or large tool outputs). Slims the on-disk rollout in place - dropping superseded compaction checkpoints, replacing images with text, truncating oversized outputs, archiving the original first - so the session reloads and can be continued. Trigger on phrases like "Codex crashes on startup", "Invalid string length", "Codex won't open my conversation", "Codex session too big", "oversized rollout", or "recover my Codex session". Not for routine context-window pressure (use Codex's /compact) and not for Claude Code sessions, which store attachments by reference and do not hit this crash.

Ship repo changes through GitHub PR review, Replit sync/checks, production publish, live verification, and sync reporting.

Operate the native Replit Desktop app through UI automation, including navigating tabs and panes, opening Preview, Database, Git, Shell, Publishing, Agent, Settings, and task board surfaces, using Replit workspace shortcuts, and handling Replit sync or publish UI flows. Use this whenever the user asks to drive Replit Desktop, inspect a Replit workspace UI, open a Replit pane or tab, use the Replit Agent panel, run workspace checks in the Replit shell, sync Git in Replit, or publish from the native app.

Download and back up Google Drive files locally via the gws CLI with format conversion and AI-readable output. Google Docs to Markdown (with image extraction), Google Sheets to CSV per tab plus XLSX, Google Slides to PPTX. Also converts downloaded .docx to Markdown and .xlsx to CSV per tab for AI readability. Supports personal Drive, shared drives, and folder structures. Use when (1) user asks to back up, download, copy, sync, or mirror Google Drive files locally, (2) user asks to export or convert Google Docs, Sheets, Slides, .docx, or .xlsx to AI-readable formats, (3) user wants offline access to Drive content, (4) user mentions drive backup, download from Drive, export Drive, or copy Drive files.

Google Chat: Manage Chat spaces and messages.

Google Drive: Manage files, folders, and shared drives.

Subscribe to Google Workspace events.

Run BellaAssist co-design archetypes as virtual feedback agents for brand evaluation, pricing debates, feature prioritisation, and messaging testing. This skill should be used when the user mentions "archetypes", "virtual panel", "brand feedback", "archetype debate", "what would the SCs think", "what would participants think", "what would Bella think", "run the panel", "archetype feedback", "brand coherence", or wants feedback on BellaAssist brand, product, or positioning from simulated user perspectives. Supports SC panel (4 B2B archetypes), Participant panel (4 end-user archetypes), Org panel (1 brand-identity archetype), full 9-voice panel, or specific archetype pairs and named debate presets.

Create user archetypes from research data or product context. Use when the user wants to build, define, or refine user archetypes, behavioural user types, or abstract user profiles. Also triggers when user mentions "archetype", "user types", or wants a non-persona approach to user segmentation. Guides a multi-step workflow from scoping through data intake, pattern identification, drafting, validation, and deliverable generation.

Process raw Co-Design Program transcripts from the Google Drive Staging_Transcripts folder into Golden Tier format. Use when asked to clean transcripts, promote staging files, or update the Golden Tier collection.

Download and back up Google Drive files locally via the gws CLI with format conversion and AI-readable output. Google Docs to Markdown (with image extraction), Google Sheets to CSV per tab plus XLSX, Google Slides to PPTX. Also converts downloaded .docx to Markdown and .xlsx to CSV per tab for AI readability. Supports personal Drive, shared drives, and folder structures. Use when (1) user asks to back up, download, copy, sync, or mirror Google Drive files locally, (2) user asks to export or convert Google Docs, Sheets, Slides, .docx, or .xlsx to AI-readable formats, (3) user wants offline access to Drive content, (4) user mentions drive backup, download from Drive, export Drive, or copy Drive files.

Generate sequential PRD prompt sequences optimized for Replit Agent development. Use when: building applications with Replit Agent, creating PRDs or technical specifications for Replit, experiencing Replit Agent failures and needing better prompting strategy, or decomposing complex projects for Replit Agent. Triggers on phrases like "build with Replit", "Replit Agent", "Replit PRD", "vibe coding".

WCAG 2.2 AA accessibility compliance for React Native and Next.js applications serving users with disabilities. Use when: (1) Building accessible mobile apps (iOS/Android), (2) Building accessible web apps, (3) Implementing accessible authentication flows, (4) Creating accessible forms or multi-step wizards, (5) Adding screen reader support, (6) Implementing keyboard navigation, (7) Meeting NDIS or disability service accessibility requirements, (8) Testing accessibility compliance, (9) Creating Easy Read or cognitively accessible content, (10) Supporting assistive technologies (VoiceOver, TalkBack, NVDA, switch access, voice control)