Claude Skills

The `review` skill performs an adversarial, evidence-based critique of a technical specification before implementation begins. It acts as a skeptical senior reviewer, attempting to refute the spec by citing concrete evidence from the codebase or research, and concludes with a clear go/no-go decision. Use this skill for high-risk projects by triggering it with phrases like "review the spec" or "red-team this."

The research skill gathers external knowledge about libraries, APIs, or best practices when the agent is uncertain, ensuring specs are based on facts. It logs each sourced finding into a durable research log (§R) and explicitly flags any unsourced claims. It triggers automatically on unclear spec decisions or via user commands like "research this" or "/ck:research".

The grill skill helps developers refine vague ideas by asking focused questions before writing a spec. It captures each answer as a clear goal (§G) or constraint (§C), parking unknowns as `?` items without guessing. Use it when you have a fuzzy concept to stress-test assumptions and prevent costly mistakes early in the planning process.

The `deepen` skill is an optional refactoring pass that identifies shallow modules and proposes design improvements to create deeper abstractions with smaller interfaces. It maintains existing behavior and test passes while suggesting explicit edits to implementation, verification, and test sections. Developers can trigger it with phrases like "deepen this" or "improve the design" to strategically use spare usage for codebase quality.

This Claude Skill provides comprehensive UI/UX design assistance for web and mobile development across 10 popular tech stacks. Use it to plan, build, review, or optimize interfaces with access to extensive design systems, guidelines, and component libraries. It helps with everything from color palettes and typography to implementing specific styles like glassmorphism or building complete projects like dashboards and e-commerce sites.

The `anydesign` skill analyzes visual inputs like images, URLs, or Figma files to extract and document their design system. It generates a structured `design.md` file containing design tokens, a component inventory, and reconstruction notes. Use it to understand, replicate, or audit the design of any visual asset, from full pages to single UI elements.

This skill provides access to Expo's official repository of ~70 version-matched integration examples for libraries like Stripe, Clerk, and Supabase. Use it when you need the canonical implementation pattern for adding a third-party service to an existing Expo app or to scaffold a new project from an example. It can fetch and explore these examples directly through the GitHub API and CLI tools.

Memex provides cross-session shared memory for agents, allowing them to persist facts and reasoning patterns via a six-layer reconstruction trail. Developers should use it at session start to load prior context, mid-session to search memory, and immediately to log new observations. Always query memex before re-deriving architectural decisions, as re-derivation indicates a gap in the memory trail.

The `memex-wrap` skill automates the handoff process after completing a milestone slice in a memex repository. It updates key documentation (`CONTINUE_HERE.md` and `ROADMAP.md`) to reflect the current state and next steps, and proposes the appropriate git tag and commit message. Use it before committing to ensure the project's trail and scoreboard are synchronized with the shipped work.

The `memex-verify` skill runs a local pre-commit gate for the memex repository, performing format checks, strict clippy analysis, and the unit test suite to mirror CI. Use it before committing changes to catch failures early, especially after editing Rust crates. It can optionally run the integration test suite if the Postgres container is active, but it is a verification tool, not a replacement for the authoritative CI.

This skill logs reasoning patterns and biases (like anchoring or confirmation bias) to a memex store for later review. It captures observations mid-session or at the end, piping them to `memex add` and running an extractor to make them queryable. It also provides a secondary path to append observations to a markdown file.

The `memex-init` skill reconstructs working context at the start of a session by loading the bias-log, reading the persistent-self trail, running the verification scoreboard, and surfacing the next documentation slice. Use it when beginning fresh work, reconstructing context, or re-orienting after an interruption. It performs the session initialization ritual, not the `memex init` CLI command for store/db setup.

Memex provides cross-session memory for agents, allowing them to persist facts and reasoning patterns via a six-layer reconstruction trail. Developers should use it to load prior context at session start, search stored observations mid-session, and immediately log new insights to avoid re-deriving past decisions.

This skill automates the handoff process after completing a milestone slice in a memex repository by updating key documentation files (`CONTINUE_HERE.md` and `ROADMAP.md`). It ensures the project state is accurately reflected, ticks off completed items, and proposes appropriate git commit/tag formatting. Use it before committing to maintain a clean trail for the next development session.

The `memex-verify` skill runs a local pre-commit gate for the memex repository, performing format checks, strict clippy analysis, and the unit test suite to mirror CI. Use it before committing changes to Rust crates or when debugging CI failures locally. It can optionally run integration tests if the Postgres container is active, serving as an early failure catch rather than a CI replacement.

This skill logs bias or reflection observations to a memex store during a session. It triggers when you detect reasoning patterns like anchoring or need to persist insights from meditation. It primarily adds entries via `memex add` for queryability and secondarily appends to an OBSERVATIONS.md file.

The `memex-init` skill reconstructs working context at the start of a session by loading the bias-log, reading the persistent-self trail, running the verification scoreboard, and surfacing the next documentation slice. Use it when beginning fresh work, reconstructing context, or re-orienting after an interruption. It is a session initialization ritual, distinct from the `memex init` CLI command that sets up the store/database.

Memex is a cross-session memory system for agents that provides three key tools: loading a persistent bias-log at session start, searching prior context mid-session, and logging new observations. Its six-layer processing trail stores both facts and the agent's own reasoning patterns across multiple sessions. Developers should query Memex to avoid re-deriving prior architectural decisions, as such re-derivation indicates a gap in the memory trail.

This skill automates the handoff process after completing a milestone slice in a memex repository by updating key documentation files (`CONTINUE_HERE.md` and `ROADMAP.md`). It ensures the project state is accurately reflected, ticks off completed items, and proposes appropriate git tags and commit messages. Developers should use it before committing to maintain a clean trail for the next session or when documentation has drifted from the shipped code.

The `memex-verify` skill runs a local pre-commit check for the memex repository, performing format checks, strict clippy linting, and unit tests to mirror CI. Use it before committing changes to Rust crates to catch failures early. It can optionally run integration tests if the Postgres container is available, but it's a verification gate, not a replacement for the official CI.

This skill logs bias or reflection observations to a memex store for persistence across sessions. It triggers when you detect reasoning biases, verification gaps, or valuable reflections mid-session. The primary method pipes content to `memex add` for queryable storage, while a secondary path appends to an OBSERVATIONS.md file.

This skill runs a session initialization ritual to reconstruct working context for the memex project. It loads the bias-log, reads the persistent-self trail, runs verification checks, and surfaces the next documentation slice. Use it at the start of any memex session or when re-orienting after an interruption.

The `create-workflow` skill helps developers author reusable, self-contained `.mjs` orchestration scripts for Claude Code's Workflow tool. It guides you through creating parameterized workflows that coordinate multiple agents with fixed control flow, auditability, and verification fail-safes. Use this skill when you need to capture a repeatable procedure as a reusable artifact instead of using a team.

The `create-workflow` skill helps developers build reusable, parameterized agent orchestration scripts (`.mjs` files) for Claude Code's Workflow tool. It guides you through the entire creation process, including template structure, metadata, fail-safes, and validation. Use this skill when you need to codify a repeatable multi-agent procedure into a fixed, auditable artifact.