Multi-model AI orchestration where behaviour is Markdown, not code. One coordinator routes scoped task packets to 71 role-based specialists across 5 model families (Codex / Claude / Gemini / Grok / Kimi) — each working in an isolated git worktree, each reviewed by a rival family before it lands. One tmux session, no server.
This repository also configures its own agents. See what claude-vibe-squad tells them →
Skill Claude CodeCodex
Use when a log, build transcript, generated artifact, or dump is too large to read whole and its middle is repetitive—sample bounded first/last slices, report the unread gap, and capture producer status separately. Not for source, briefs, packets, or configuration; their whole-file rule comes from the injected task…
Skill Claude CodeCodex
Use when a suspected or active security compromise must move through evidence-preserving triage, containment planning, eradication, recovery, and post-incident review while every live action remains separately operator-gated. Not for an ordinary reliability outage or chronology reconstruction alone.
Skill Claude CodeCodex
Use when gameplay state must drive sound and music reactively—specify stems, crossfades, stingers, horizontal resequencing, sound-effect pools, round-robin variation, spatialization, bus ducking, voice limits, and memory or streaming ceilings. Stop at the reactive-system blueprint; downstream schema packaging, media…
Skill Claude CodeCodex
Use before implementing against an API, schema, file format, or contract you did not author, especially when units, null/empty/absent behavior, ordering, errors, retries, or idempotency are unstated—pin each load-bearing assumption with documentation, observed producer/consumer behavior, or a probe. Use requirements…
Skill Claude CodeCodex
Use when supplied or grounded search queries must be grouped into reproducible intent/topic clusters and mapped one-cluster-per-page; emit the evidence table with source/date/locale/device and explicit split/merge rules. Not for measuring ranking impact without analytics.
Skill Claude CodeCodex
Use when implementing or evaluating a retrieval-augmented product agent that must enforce corpus ACLs, trace answers to returned passages, fail safely on no-hit, and pass retrieval, injection, and refresh regressions. Not for general repository memory or chrono-vault recall.
Skill Claude CodeCodex
Use when analyzing a target deeply enough to require distinct structural, behavioral, edge-case, and contradiction passes rather than one undirected read—fix a question for each layer, record its evidence-backed delta, and stop only at a no-delta layer or an explicit budget cutoff.
Skill Claude CodeCodex
Use when an upstream mechanics brief must become playable spatial structure—block out spaces and critical paths, place gates and rewards, shape pacing and challenge, map each beat to its intended feeling, and define per-stage playtest assertions. It owns layouts and encounter flow; upstream system tuning and…
Skill Claude CodeCodex
Use when content must be rewritten for a specific region or language while preserving intent and tone—adapt idiom, dialect, register, cultural references, and protected glossary terms; flag meanings that cannot transfer cleanly. Not for literal word-for-word translation.
Skill Claude CodeCodex
Use when gameplay needs a story arc, branching quest graph, character relationships, or dialogue outline that reinforces the player fantasy—assign stable node IDs, resolve prerequisites, align beats to play, and mark narrative choices that change mechanics. It owns story and quest causality; use level-design-patterns…
Skill Claude CodeCodex
Use when a finished or near-finished media asset must be exported for a named publishing platform—retrieve current official resolution, aspect, duration, codec, loudness, caption, safe-area, overlay, and disclosure requirements, then verify the actual render against a per-platform checklist. It checks delivery-channel…
Skill Claude CodeCodex
Use when a game's motivation, reward cadence, or retention loop is being designed or reviewed for engagement and ethical risk—tie choices to autonomy, competence, relatedness, and flow; document variable rewards and flag FOMO, predatory monetization, or addictive loops. It evaluates motivational psychology, not core…
Skill Claude CodeCodex
Use when an operator or stakeholder request is too vague or conflicting to convert directly into testable scope: ask a bounded question ladder, label assumptions DEFAULT or BLOCKING, turn adjectives into measurable outcomes, and confirm IN and OUT boundaries plus verification methods before decomposition.
Skill Claude CodeCodex
Hard Rule 6 pre-publication rights gate — check licensing/provenance AND likeness/voice consent over a generated or third-party media asset before publish/use, and emit the machine-readable gate record. Use at the S4 Verify step of content/image, content/video, content/audio-assets (or any asset headed for…
Skill Claude CodeCodex
Use at S4 before an investigation, article, campaign, narration, or screen production with external-world assertions may advance: require returned source bundles, enforce the requested recency interval for every decision-critical premise, reject unmapped items, and emit the mandated PASS/needstool truth record.
Skill Claude CodeCodex
Use after the goal is understood but the work spans multiple modules, owners, or verification seams: cut it into dependency-ordered, independently shippable units with explicit write sets, exclusions, and one-sentence pass or fail checks.
Skill Claude CodeCodex
Use before analyzing any corpus to determine how much can actually be read and how strong later coverage claims may be—measure file count, total bytes, and largest item; name the binding context, tool-call, or time budget; assign full-read, sample, and index-only tiers; and carry numeric denominators into the…
Skill Claude CodeCodex
Use when security findings or incident responsibilities are stalling because component ownership and decision rights are unclear—derive an evidence-backed map from code ownership, commit, deploy, and on-call records; record patch, shutdown, and credential-rotation authority; and flag unowned, departed, shared, vendor…
Skill Claude CodeCodex
Use before designing or reviewing a system whose assets, adversary, trust zones, and residual risks are not yet explicit: enumerate reachable STRIDE threats at each boundary, distinguish observed controls from assumptions, and terminate at a declared reporting floor.
Skill Claude CodeCodex
Use after a defect pattern is confirmed and you need a reusable Semgrep detector for sibling instances: derive syntactic or taint rules from vulnerable and safe fixtures, encode project sanitizers, and measure repository-wide false positives. Not for choosing generic rulesets or scanning only a change.
Skill Claude CodeCodex
Use when a description frontmatter field must become an accurate activation trigger, especially because its skill stays dormant or is confused with a sibling—encode recognizable workflow cues and replay-test positive and negative tasks. Scope is trigger metadata, not instruction-body or resource design.
Skill Claude CodeCodex
Use when a UI, film, or game needs individual sounds or a coherent sonic family designed to communicate an event, affordance, feedback, or mood—specify source and processing, timing envelope, differentiation, and target playback context while routing recognizable samples or voices through rights review. For reactive…
Skill Claude CodeCodex
Use when creating or correcting JSON-LD or schema.org markup for a page and the chosen types, required properties, and values must match visible content and validate without cloaking or spam.
Skill Claude CodeCodex
Use when auditing whether anything outside the project can silently alter a build or published release—inventory dependencies, base images, binaries, generators, CI actions, install hooks, secret exposure, and publisher controls; verify pins, digests or signatures, reproducibility, and one artifact's end-to-end…
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: