Skill Claude CodeCodex
Multi-pass, multi-lens security review that reinforces detection through independent blind finder passes and an adversarial refute pass. Use when reviewing an owned or authorized codebase/change for vulnerabilities with high confidence, or when a change touches auth, data, crypto, external input, dependencies, or…
Skill Claude CodeCodex
Authorization-gated white-hat penetration testing against an owned or explicitly authorized target. Follows find → validate → fix → regress. Requires the authorization gate before any active testing (crafted requests, exploit runs, scanning, fuzzing). Use for "pentest my app", "test this endpoint for vulnerabilities"…
Skill Codex
Find and load the right Codex skill from a large local library when skill descriptions are shortened, hidden, explicit-only, or no visible skill clearly matches. Use before giving up on skill context, after skill or plugin library changes, or when Codex needs smart access to all installed skills without injecting…
Command Claude Code
Act as a UX/product partner in two modes — design-maker (design in the live mockup, keep one source of truth, run a design signoff, propagate signed-off UI to Figma and the board) OR prototype-consumption/backlog-shaping (consume an existing prototype/Figma/screenshots/repo, inventory…
Skill Codex
UX/product workflow for design-makers and consumers of existing designs. Use for making or revising mockups, screens, layouts, flows, prototypes, tokens, typography, components, design systems, live design previews, signoff, and Figma or board handoff. Also use to consume a prototype, Figma file, screenshots, or UI…
Skill Claude CodeCodex
Use when writing or debugging a bb plugin — imports failing at load, a migration corrupting state, an RPC call rejected before it leaves the browser, or a build that works in a shell and not in bb.
Skill Claude CodeCodex
Use when calling the bb plugin SDK — listing providers or threads, reading thread events, spawning threads on a machine, or creating worktrees. Several calls return confident, wrong-looking-right answers; this is what each one actually does.
Skill Claude CodeCodex
Use when a coding agent on a headless Linux box or VPS FAILS — every shell command denied, "setting up uid map: Permission denied", a sandbox or bwrap error, a login that hangs with no code printed, EACCES on a global npm install, or a daemon that dies when SSH disconnects. Host policy masquerading as agent bugs.
Skill Claude CodeCodex
Use when building or debugging token/cost accounting across agent providers, or when a spend dimension renders empty. Covers what each event source can and cannot tell you, why skill usage is nearly invisible, and how to bucket a usage chart so it is readable at every window.
Skill Claude CodeCodex
Use when running a fleet of agents, or when the same class of mistake keeps recurring across sessions. Failure patterns that repeat, why each one survives review, and the check that catches it. Read before declaring work done.
Skill Claude CodeCodex
Use when sending work to another agent for review, critique or validation, or when following up on a review that already happened. Covers why a re-review must continue the original conversation, and what a reviewer needs in order to answer instead of flailing.
Skill Claude CodeCodex
Use before writing, replacing, or publishing QA instructions in Jira, Linear, or another tracker; before moving work to Ready for QA or calling interface end-to-end testing complete or blocked; and whenever a tester journey depends on login, seeded accounts, role-specific portals, generated links, tokens, permissions…
Skill Claude CodeCodex
Use after building or changing any chart, dashboard or visual layout, and when a user says a chart "isn't working" but the data looks right. Explains why a text snapshot cannot tell you a chart rendered, and what to measure instead.
Skill Codex
Check that a plan, delegated brief, or ongoing result covers the complete user request without adding unrelated work. Use for substantial ambiguity, complex handoffs, material scope revisions, suspected instruction drift, or closeout of multi-part work.
Skill Codex
UI/UX design intelligence for web, mobile, and desktop. This skill should be used when designing, building, reviewing, or fixing interfaces, including pages, components, design systems, accessibility, interaction, responsive layout, typography, color, charts, and stack-specific UI implementation. Searchable local…
Skill Claude CodeCodex
Use whenever an agent step is a DECISION rather than writing — routing, triage, in scope or not, risky or not, severity, pass/fail, done or not, escalate or not, which model or lane, whether a finding is real. Also use when designing a gate, a verdict, a review outcome, a delegate's return contract, or a hook that…
Skill Claude CodeCodex
Use before starting expensive, release, migration, or otherwise single-flight agent work that must not run concurrently for the same repository, worktree, branch, commit, and operation. Acquires a fail-closed host lease with owner metadata, heartbeat renewal, deterministic stale handling, and release.
Skill Claude CodeCodex
Use before marketing copy goes in front of the brand owner or the public, for a landing hero or section, ad, post, pricing block or email subject for a brand that has a profile, especially Brazilian mental-health marketing aimed at psychologists and their patients. Returns a typed accept, verify or revise decision…
Skill Claude CodeCodex
Use in repos with a committed .qa/config.json before shipping (a protected push, a release, any production deploy), and whenever a workflow walk finds defects. Covers the full pipeline the ship-gate enforces: scope the workflow (P0), walk it end to end and inventory EVERY defect without fixing anything (P1, including…
Skill Claude CodeCodex
Keep a coordinator on what the user asked for. A per-thread ledger of the user's purposes (verbatim), a PreToolUse gate that denies spawns and tells that do not say which open purpose they serve, a Stop check that blocks an early stop while open purposes have nothing running, and the fleet idle guard that nudges a…
Skill Claude CodeCodex
Use when fixing a bug, touching shared code, configuration, hooks or skills, encountering a blocker, or preparing to stop. Complete the authorized user outcome, repair known defects, check consumers, and distinguish implementation from installation and exercised workflows.
Skill Claude CodeCodex
Use before selecting or reporting tests, claiming a feature or workflow complete, or exercising a product as its user. Require a named persona, identified target, stated user outcomes, and a verdict per outcome; preserve evidence boundaries and fix authorized workflow failures.
Skill Claude CodeCodex
Use when routing bounded or bulk execution through the verified Codex route, including choosing the accountable BB thread or fleet path, preparing a bounded brief, and reconciling worker output.
Skill Claude CodeCodex
Use in any repo, new or legacy, to decide whether a change works and to bring the repo up to standard. doctor audits what is missing (tests, journeys that walk the real app, evals for LLM output, dry runs for data-moving code, mutation testing, CI). init writes the repo's definition of done. run executes it and…
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: