Skill Claude CodeCodex
Make write-access AI trustworthy: every edit diffed, accept-gated, reversible, proportional, and willing to do nothing. Use when designing or auditing an agent that creates/updates/deletes user data, when adding overnight/batch AI passes, when a product’s AI writes feel silent or over-eager, or when the user asks for…
Skill Claude CodeCodex
Keep multi-platform products behaviorally one product: write platform-neutral behavior contracts per form/primitive, let each kit satisfy them natively, audit contract drift (not screenshots). Use when adding a second surface (web/iOS/macOS), when parity tickets pile up, when lightboxes/composers fork, or the user…
Skill Claude CodeCodex
Author and govern a product North Star (constitution) that every feature is judged against — premise, non-negotiables, creed, decision checklist, conflict rule. Use when starting a non-trivial product/tool, when agents disagree about what the product is, when the user asks for a manifesto / constitution / north star /…
Plugin Claude Code
Plugin marketplace listing 1 plugin: linear-methodology.
Skill Claude CodeCodex
Inventory-before-rewrite: map every table/endpoint/engine/UI to a destination tag, prove zero regressions, retire only when replacement is live. Use when migrating/reframing an existing product, consolidating AI engines, or the user asks for an inventory audit / lossless migration / “don’t lose anything.” NOT for…
Skill Codex
Owns complex multi-agent delivery as the sole root orchestrator. Use when the user says “orchestrate,” “act as the orchestrator,” “own the complete buildout,” “spin up subagents,” “manage agents until done,” or asks one root agent to own readiness, task progression, worker challenge, integration, and final delivery.…
Skill Claude CodeCodex
Write or review clear commit messages, PR titles, and PR descriptions that explain what changed and why. Use before writing git history or when asked to improve it. Not for code comments, chat replies, or docs pages.
Skill Codex
Give sparse, evidence-based progress updates for long-running agent tasks and projects. Use when work will take more than 15 minutes, spans multiple milestones, involves delegated agents, or when the user asks for progress bars, status cadence, fewer updates, or a global cross-agent progress rule. Not for short…
Skill Claude CodeCodex
Use when sharing in-progress UI, product, or social screens for approval. Composes 1 desktop or 1–3 mobiles onto a dotted mat. Not for marketing heroes, App Store shots, final production, or raw Figma dumps.
Skill Claude CodeCodex
Build, adapt, or review native iOS and iPadOS apps with Apple frameworks and selectively retrieved components, interaction recipes, and verification guidance from native.offbr.co. Use for SwiftUI product work, web-to-native migrations, and native UI audits.
Skill Claude CodeCodex
Route routine agent work to deterministic tools and bounded JEV decisions through the user’s preferred connection. Use when the user requests an executive assistant, EA, cheaper task delegation, or supervised computer-use assistance.
Skill Claude CodeCodex
Keep automated testing proportional to concrete risk and maintenance cost. Use when deciding whether to add tests for a feature, bug fix, or refactor, or when asked to reduce unnecessary testing.
Skill Claude CodeCodex
Make a product launch film by default, or a focused demo when requested, from an app, package, feature, or website. Show the real product UI and motion, then render and mix the film.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: