If your AI supports skills, VibeSkills works. 340+ skills spanning coding, research, data science & creative work.
π§ Planning Β· π οΈ Engineering Β· π€ AI Β· π¬ Research Β· π¨ Creation
Install Β βΒ vibe | vibe-want | vibe-how | vibe-do Β βΒ Smart Routing Β βΒ M / L / XL Execution Β βΒ Governance Verification Β βΒ β Delivery
- What makes it different
- Who is it for
- Intelligent Routing
- Memory System
- Full Capability Map
- Installation & Management
- Getting Started
π New here? Quick glossary of key terms (click to expand)
| Term | Plain-English Meaning |
|---|---|
| VibeSkills / VCO | This project. VCO = Vibe Code Orchestrator β the runtime engine behind the skills. |
| Skill | A focused capability module (e.g., tdd-guide, code-review). Think of skills as expert assistants the system calls on demand. |
| Governed runtime | When you invoke vibe, the system follows a structured process β clarify β plan β execute β verify β instead of diving in blindly. The public discoverable wrapper set is vibe, vibe-want, vibe-how, and vibe-do; hosts may render them as labels like Vibe: What Do I Want?, but they still resolve to the same canonical runtime authority. |
| Canonical Router | The internal logic that decides which skill to activate for your task. Just invoke /vibe and let it route automatically. |
| M / L / XL grade | Task complexity level. M = quick focused task, L = multi-step task, XL = large task with parallel work. Automatically selected. Public overrides are limited to --l and --xl; they are execution preferences, not separate entrypoints. |
| Frozen requirement | Once you confirm the plan, it is "frozen" β the system will not silently change scope mid-task. |
| Root / Child lane | In XL tasks, there is a "root" coordinator and "child" worker agents. Prevents conflicting outputs from parallel agents. |
| Proof bundle | Evidence that a task was actually completed correctly β test results, output, verification logs. |
Important
VibeSkills evolves with the times β ensuring it stays genuinely useful while dramatically lowering the barrier to cutting-edge vibecoding technology, eliminating the cognitive anxiety and steep learning curve that comes with new AI tools.
Whether or not you have a programming background, you can directly harness the most advanced AI capabilities with minimal effort. Productivity gains from AI should be available to everyone.
Traditional skill repos answer: "What tools do I have?" VibeSkills tackles the core pain point of heavy AI users: "How do I manage and invoke large numbers of Skills, and get work done efficiently and reliably?"
Works with Claude Code Β· Codex Β· Windsurf Β· OpenClaw Β· OpenCode Β· Cursor and any AI environment that supports the Skills protocol. Native MCP compatibility.
| β Β Traditional Pain Points (you've probably felt these) | β Β VibeSkills Solutions (what we've built) |
|---|---|
| Skills never activate: Hundreds of capabilities in the repo, but AI rarely remembers to use them β activation rate is extremely low. | π§ Intelligent Routing: The system automatically routes to the right skill based on context β no need to memorize a skill list. |
| Blind execution: AI dives in without clarifying requirements β fast but off-target, projects gradually become black boxes. | π§ Governed Workflow: Clarify β Verify β Trace is enforced in a unified process; every step is auditable. |
| Conflicting tools: Lack of coordination between plugins and workflows leads to environment pollution or infinite loops. | 𧩠Global Governance: 129 contract rules define safety boundaries and fallback mechanisms for long-term stability. |
| Messy workspace: After extended use, repos become cluttered; new Agents miss project details when taking over, causing handoff gaps. | π Semantic Directory Governance: Fixed-architecture file storage so any new AI conversation instantly understands the project context. |
| AI bad habits: Deletes main files while clearing backups; writes silent fallbacks then confidently claims "it's done". | π‘οΈ Built-in Safety Rules: Governed execution blocks dangerous bulk deletion and blind recursive wipes by default; fallback mechanisms must always show explicit warnings. |
| Manual workflow discipline: Users must maintain their own AI collaboration process from experience β high learning cost. | π¦ Framework-guided end-to-end: Requirements β Plan β Multi-agent execution β Automated test iteration β fully managed. |
| Skill dispatch chaos in multi-agent runs: Hard to assign the right skills to each agent for different tasks. | π€ Automatic Skill Dispatch: Multi-agent workflows automatically assign the corresponding Skills to each Agent's task. |
Which of those pain points hit home? Find your position β what comes next will land harder.
Is this for you? Click to expand
| Audience | Description |
|---|---|
| π― Users who need reliable delivery | Want AI to be a dependable partner, not a runaway horse |
| β‘ Power users heavily relying on AI/Agents | Need a unified foundation to support large-scale workflows |
| π’ Small teams with high standardization needs | Want AI workflows to be more maintainable and transferable |
| π© Practitioners exhausted by skill sprawl | Already tired of tool hunting β just want a ready-to-use solution |
If you're looking for a single small script, this may be overkill. But if you want to use AI more reliably, smoothly, and sustainably β this is your indispensable foundation.
You know this is for you. Next question: 340+ skills in one system β how do they stay out of each other's way?
With 340+ skills, you might wonder: "Won't similar skills conflict? How does the system know which one to use?"
VibeSkills uses a Canonical Router as the single authoritative routing decision center:
graph LR
A[User Task] --> B{Canonical Router}
B --> C[Intent Recognition]
C --> D[Keyword Extraction]
D --> E[Skill Matching]
E --> F[Conflict Detection]
F --> G[Priority Ranking]
G --> H[Routing Decision]
H --> I[Execute Skill]
style B fill:#7B61FF,stroke:#fff,stroke-width:2px,color:#fff
style F fill:#FF9800,stroke:#fff,stroke-width:2px,color:#fff
VibeSkills follows a Clarify β Plan β Execute β Verify governed workflow to ensure every task goes through complete quality control:
- Requirements Clarification: Skills like
speckit-clarifydefine clear boundaries and acceptance criteria - Architecture Planning: Skills like
aios-architectdesign the implementation path - Execution Layer: 340+ skills called on demand to complete the actual work
- Quality Verification: Skills like
tdd-guideandcode-reviewensure delivery quality
Traditional skill repos let AI "freely choose" β the result:
- β AI can't remember what skills exist
- β Similar skills conflict with each other
- β Execution paths are unpredictable
VibeSkills routing guarantees:
- β Determinism: Same task always follows the same routing logic
- β Traceability: Every routing decision has a clear rationale
- β
Control: Users can override default routing via explicit invocation (e.g.
/vibe) - β Stability: 129 governance rules prevent conflicts and divergence
After selecting the primary skill, the router also automatically determines the execution level based on task complexity:
| Level | Use Case | Characteristics |
|---|---|---|
| M | Narrow-scope work with clear boundaries | Single-agent, token-efficient, fast response |
| L | Medium complexity requiring design, planning, and review | Native serial execution by planned steps; bounded delegated units only when explicitly planned |
| XL | Large tasks β parallelizable, long-running, multi-agent wave execution | Wave-sequential orchestration with step-level bounded parallelism for independent units only |
The system automatically selects the level after requirements clarification, before plan execution. Public host-visible entry wrappers ship as
vibe,vibe-want,vibe-how, andvibe-do. Hosts may render them asVibe,Vibe: What Do I Want?,Vibe: How Do We Do It?, andVibe: Do It, but they still route into the same governed runtime.When the system calls a specialist skill internally (like
tdd-guideorcode-review), it is always scoped to a specific phase β they assist without taking over the overall coordination. In XL tasks with multiple agents, worker agents (child lanes) can suggest specialist help, but the coordinator (root) approves it before execution.You can also express an explicit preference:
Please execute this task according to the plan, launching XL-level workflow /vibe
The only lightweight public grade overrides are
--land--xl. Aliases likevibe-l,vibe-xl, orvibe-how-xlare intentionally unsupported.
π Routing FAQ (click to expand)
One route or multiple per task?
Core principle: A task typically routes to one primary skill, but that skill can invoke others as sub-processes.
- Single primary route: The Canonical Router selects the single best-matching primary skill
- Skill composition: The primary skill can invoke others as needed during execution (e.g.
vibecan invokespeckit-clarify,aios-architect, etc.) - Governed coordination: Multi-skill collaboration is controlled by governance rules, not arbitrary combinations
How are conflicts between similar skills handled?
When multiple skills appear capable of completing a task, the router avoids conflicts through:
- Priority rules: Each skill has a clear priority and applicable scenario
- Context matching: Analyzes task complexity, multi-phase needs, and explicit user preferences
- Mutual exclusion rules: 129 rules include exclusion rules preventing conflicting combinations
- Graceful degradation: When the preferred skill is unavailable, fallback by priority β no infinite loops
Will too many options cause token explosion?
No. Routing doesn't dump all options into the model β it uses a smart trigger mechanism:
User command β AI-assisted governance extracts intent keywords β keywords trigger skill routing
The governance framework adds ~30k initial context overhead, but does not cause token explosion.
Real example: User says "Help me refactor this project"
- Intent recognition β Complex refactoring task
- Keyword extraction β refactor, project, code quality
- Skill matching β
vibe/autonomous-builder/systematic-debugging - Routing decision β Choose
vibe(refactoring needs multi-phase: clarify β plan β execute β verify)
Routing solves "which skill". But there's a deeper question: when the conversation ends, does AI remember you?
Sound familiar?
| β Pain Point | β VibeSkills Solution | Component |
|---|---|---|
| Re-explaining project context every new session | Architecture decisions & conventions auto-loaded on startup | Serena |
| AI hits the same bugs again; insights vanish with context | One sentence saves to Obsidian + GitHub permanently | knowledge-steward |
| Long tasks β AI gradually "forgets" early context | In-session semantic vector cache, instant retrieval | ruflo |
| Cross-project knowledge can't accumulate | Entity relationship graphs grow richer over time | Cognee |
| Long task interrupted β hard to hand off to new agent | Auto-folds into working + tool + evidence memory | deepagent-memory-fold |
π Expand: Four-Tier Architecture, Memory Skills & Governance Rules
VibeSkills builds a four-tier memory system β one authoritative component per memory need:
| Tier | Component | Scope | Core Purpose |
|---|---|---|---|
| L1 Session | state_store |
Current session | Execution progress, intermediate results, temp state β always-on "workbench" |
| L2 Project | Serena |
Current project | Architecture decisions, conventions β written only after explicit user confirmation |
| L3 Short-term Semantic | ruflo |
Intra-session | Vector cache for fast context retrieval within long-running tasks |
| L4 Long-term Graph | Cognee |
Cross-session | Entity linking, relationship graphs, long-horizon knowledge accumulation |
Optional extensions:
mem0as a personal preference backend (opt-in);Lettaprovides memory block mapping vocabulary β neither replaces the four canonical tiers.
Three Dedicated Memory Skills
| Skill | Role | Trigger |
|---|---|---|
knowledge-steward |
Knowledge Keeper: Saves insights, bug fixes, and prompts to Obsidian + GitHub permanently | "save this prompt" / "log this bug" / "save this insight" |
digital-brain |
Second Brain: Structured personal knowledge base β identity, content, network, retrospectives | Invoke directly; ideal for a personal knowledge OS |
deepagent-memory-fold |
Context Fold: Compresses large context into structured working/tool/evidence memory for seamless handoff | Triggers at context limit or manually |
Governance: Single source of truth (no dual-track) Β· Explicit write only (Serena requires confirmation) Β· episodic-memory permanently disabled Β· mem0 limited to personal preferences Β· Kill switch on every external backend
Routing + memory form the dispatch nervous system. Here's the full capability map they power β end to end.
Unrolled across a "real workflow", VibeSkills has laid out a complete end-to-end capability chain:
| Domain | Coverage | Representative Engines |
|---|---|---|
| π‘ Requirements & Clarification | No more black-box starts: turn vague ideas into clearly-bounded, verifiable problem definitions | brainstorming, speckit-clarify |
| π Planning & Breakdown | Decompose ambitious goals into specs, plans, tasks, milestones, and execution flows | writing-plans, speckit-specify, aios-po |
| ποΈ Architecture & Tech Selection | Design frontend/backend boundaries, APIs, data layers, deployment, and tech stack comparisons | aios-architect, architecture-patterns |
| π» Development & Implementation | New features, scaffolding, engineering integration, and precise cross-file implementation | autonomous-builder, speckit-implement |
| π§ Debugging & Refactoring | Beyond surface patches: locate errors, analyze root causes, restore project-level maintainability | error-resolver, systematic-debugging |
| π‘οΈ Testing & Quality Control | Unit tests, regression verification, quality gates β mandatory verification before completion | tdd-guide, aios-qa, code-review |
| π Collaboration & Release | Handle Issues/PRs, fix CI, process reviews, and automated deployment | aios-devops, gh-fix-ci, vercel-deploy |
| π€ Compound Workflows | Freeze requirements, dispatch tasks, multi-Agent coordination, execution tracing, env cleanup | vibe, swarm_*, hive-mind-advanced |
| π External Ecosystem | Bridge browsers, web scraping, design files, third-party services, and context memory | mcp-integration, playwright, scrapling |
| π Data & AI Engineering | EDA, cleaning & stats, to model training, RAG retrieval, and experiment tracking | senior-ml-engineer, statistical-analysis |
| π¬ Research & Life Sciences | Core strength: literature review, bioinformatics, single-cell analysis, drug discovery | literature-review, biopython, scanpy |
| π Math & Scientific Computing | Symbolic derivation, Bayesian modeling, multi-objective optimization, simulation, quantum computing | sympy, pymc-bayesian-modeling, qiskit |
| π¨ Multimedia & Presentation | Interactive charts, publication-quality figures, slides, audio/video production | plotly, generate-image, video-studio |
π Expand: Explore the complete 340+ full-stack capability matrix
π‘ Why governance matters: The vast skill library below is not a collection of isolated scripts β it is an ecosystem governed by the VCO runtime. Through domain matrix classification, the system automatically invokes the right tool at the right context node, without requiring you to manually search through skills.
Turn big ideas into actionable plans: requirement insights, problem definition, Sprint planning, task breakdown, and constraint collection. Ensure direction is clear, boundaries are defined, and milestones are verifiable before writing a single line of code.
.system, aios-pm, aios-po, aios-sm, aios-squad-creator, aios-ux-design-expert, brainstorming, create-plan, designing-experiments, planning-with-files, shared-templates, speckit-analyze, speckit-checklist, speckit-clarify, speckit-constitution, speckit-plan, speckit-specify, speckit-tasks, speckit-taskstoissues, subagent-driven-development, think-harder, treatment-plans, ux-researcher-designer, writing-plans
The true engineering foundation: from scaffolding, cross-file changes, API design to microservice architecture evaluation. Not just code output β also context memory, toolchain orchestration, and multi-phase intelligent Agent coordination.
aios-architect, aios-dev, aios-master, architecture-patterns, autonomous-builder, cancel-ralph, coding-tutor, context-fundamentals, context-hunter, cs-foundations, deepagent-memory-fold, deepagent-toolchain-plan, evaluating-code-models, get-available-resources, hive-mind-advanced, local-vco-roles, nowait-reasoning-optimizer, prompt-lookup, ralph-loop, skill-creator, skill-lookup, spec-kit-vibe-compat, speckit-implement, superclaude-framework-compat, theme-factory, vibe, webthinker-deep-research
Guarding the lifeline of code and systems: unit tests, root cause analysis, dependency conflict resolution, security vulnerability reviews, and a complete TDD guide β ensuring systems never enter a "breaks after every change" black-box state.
aios-qa, build-error-resolver, code-review, code-review-excellence, code-reviewer, data-quality-checker, data-quality-frameworks, debugging-strategies, deslop, detecting-performance-regressions, error-resolver, evals-context, experiment-failure-analysis, generating-test-reports, ml-data-leakage-guard, performance-testing, property-based-testing, providing-performance-optimization-advice, receiving-code-review, requesting-code-review, reviewing-code, security-best-practices, security-ownership-map, security-reviewer, security-threat-model, systematic-debugging, tdd-guide, verification-before-completion, verification-quality-assurance, windows-hook-debugging
Let data tell the truth: a one-stop data processing engine from data cleaning, missing value handling, and exploratory analysis (EDA) to advanced statistical testing, regression models, and time series forecasting.
aios-data-engineer, anomaly-detector, correlation-analyzer, dask, data-artist, data-exploration-visualization, data-normalization-tool, detecting-data-anomalies, excel-analysis, exploratory-data-analysis, feature-importance-analyzer, geopandas, hypothesis-testing, metric-calculator, networkx, performing-causal-analysis, performing-regression-analysis, polars, preprocessing-data-with-automated-pipelines, regression-analysis-helper, running-clustering-algorithms, scientific-data-preprocessing, splitting-datasets, spreadsheet, statistical-analysis, statistics-math, statsmodels, usfiscaldata, vaex, xlsx
Full-stack AI model development: beyond just calling APIs β feature engineering, model training, fine-tuning, interpretability (SHAP), large model evaluation (Evals), and reinforcement learning workflows.
LQF_Machine_Learning_Expert_Guide, aeon, datamol, deepchem, embedding-strategies, engineering-features-for-machine-learning, evaluating-llms-harness, evaluating-machine-learning-models, explaining-machine-learning-models, geniml, ml-pipeline-workflow, openai-knowledge, pufferlib, pytorch-lightning, scikit-learn, scikit-survival, senior-computer-vision, senior-data-scientist, senior-ml-engineer, senior-prompt-engineer, shap, similarity-search-patterns, sparse-autoencoder-training, stable-baselines3, tensorboard, timesfm-forecasting, torch-geometric, torch_geometric, torchdrug, training-machine-learning-models, transformer-lens-interpretability, transformers, umap-learn, unsloth, weights-and-biases
A formidable interdisciplinary powerhouse: single-cell sequencing analysis, protein structure folding, drug molecule discovery, genomics alignment β seamlessly integrated with cloud-based biology lab systems.
adaptyv, alphafold-database, anndata, arboreto, benchling-integration, biopython, bioservices, cellxgene-census, cobrapy, deeptools, diffdock, dnanexus-integration, esm, etetoolkit, flowio, gene-database, gget, ginkgo-cloud-lab, gtars, histolab, imaging-data-commons, labarchive-integration, lamindb, latchbio-integration, matchms, medchem, molfeat, neurokit2, neuropixels-analysis, omero-integration, opentrons-integration, pathml, protocolsio-integration, pydeseq2, pydicom, pyhealth, pylabrobot, pyopenms, pysam, pytdc, rdkit, scanpy, scikit-bio, scvi-tools, tiledbvcf
Precise derivation and complex system simulation: symbolic math, Bayesian probabilistic programming, quantum computing simulation, multi-objective optimization, and rigorous propositional logic and mathematical proof assistance.
astropy, cirq, dialectic, fluidsim, gradient-methods, math, math-model-selector, math-tools, mathematical-logic-expert, matlab, pennylane, pymatgen, pymc, pymc-bayesian-modeling, pymoo, propositional-logic, qiskit, qutip, rowan, simpy, sympy, xan
The expressway for academic productivity: precise search across dozens of databases (PubMed, arXiv, etc.), systematic review matrix organization, citation management, and the complete publication pipeline from drafting to peer review.
bgpt-paper-search, biorxiv-database, brenda-database, chembl-database, citation-management, clinical-decision-support, clinical-reports, clinicaltrials-database, clinpgx-database, clinvar-database, comprehensive-research-agent, content-research-writer, cosmic-database, datacommons-client, documentation-lookup, drugbank-database, ena-database, ensembl-database, fda-database, geo-database, gwas-database, hmdb-database, hypothesis-generation, kegg-database, literature-matrix, literature-review, manuscript-as-code, market-research-reports, metabolomics-workbench-database, open-notebook, openalex-database, opentargets-database, paper-2-web, pdb-database, peer-review, pubchem-database, pubmed-database, pyzotero, reactome-database, research-grants, research-lookup, scholar-evaluation, scholarly-publishing, scientific-brainstorming, scientific-critical-thinking, scientific-reporting, scientific-writing, string-database, submission-checklist, uniprot-database, uspto-database, zinc-database
Making knowledge and data visible: interactive chart generation, publication-quality scientific figures, slide creation, audio/video production, and deep read/write parsing of Word, PDF, and other office documents.
algorithmic-art, creating-data-visualizations, data-storytelling, datavis, doc, docs-review, docs-write, document-skills, docx, docx-comment-reply, figma, figma-implement-design, file-organizer, g2-legend-expert, generate-image, imagegen, infographics, latex-posters, latex-submission-pipeline, markdown-mermaid-writing, markitdown, matplotlib, pdf, plotly, pptx-posters, report-generator, scientific-schematics, scientific-slides, scientific-visualization, screenshot, seaborn, slides-as-code, smart-file-writer, speech, structured-content-storage, transcribe, venue-templates, video-studio, visualization-best-practices, vscode-release-notes-writer, writing-docs
Breaking the limits of the runtime: seamlessly connect external browsers, design platforms, and cloud services via MCP protocol and Playwright automation β plus CI/CD pipeline support and one-click automated deployment.
aios-devops, alpha-vantage, claude-skills, commit-with-reflection, denario, digital-brain, edgartools, flashrag-evidence, fred-economic-data, geomaster, gh-address-comments, gh-fix-ci, hedgefundmonitor, hypogenic, iso-13485-certification, jupyter-notebook, knowledge-steward, mcp-integration, modal, modal-labs, netlify-deploy, openai-docs, perplexity-search, playwright, prowler-docs, scrapling, sentry, skypilot-multi-cloud-orchestration, vercel-deploy
Now for the numbers. This isn't a demo project β it's a running system.
The runtime core behind VibeSkills is VCO. This is not a single-point tool or a "code completion" script β it is a super-capability network that has been deeply integrated and governed:
| 𧩠Skill Modules | π Ecosystem | βοΈ Governance Rules |
|---|---|---|
| Directly callable Skills covering the full chain from requirements to delivery |
Absorbed high-value upstream open-source projects and best practices |
Policy rules and contracts ensuring stable, traceable, divergence-free execution |
Skills keep growing β but you don't need to manage them individually.
The default install is intentionally narrow on the outside:
- public host-visible runtime entry:
<target-root>/skills/vibe - internal bundled specialist corpus:
<target-root>/skills/vibe/bundled/skills/* - compatibility projections: only host-scoped, explicit, and test-backed when a host still needs them
This is the key architectural split: capability breadth is internal, public surface breadth is minimal. A full install still carries broad routing and specialist coverage, but it no longer needs to flood <target-root>/skills/ with 300+ peer directories just so hosts can discover vibe.
Running uninstall.ps1 -HostId <host> or uninstall.sh --host <host> is the partner surface to install. By default it performs a ledger-first, owned-only cleanup that only touches paths recorded in .vibeskills/install-ledger.json, *.host-closure.json, or the documented legacy surfaces. The bundled runtime keeps only the executable contract; the full governance explainer lives in the canonical repo at docs/uninstall-governance.md.
The .vibeskills brand is now split into two layers on purpose:
- host-sidecar:
<target-root>/.vibeskills/host-settings.json,host-closure.json,install-ledger.json,bin/* - workspace-sidecar:
<workspace-root>/.vibeskills/project.json,.vibeskills/docs/requirements/*,.vibeskills/docs/plans/*,.vibeskills/outputs/runtime/vibe-sessions/*
This keeps host install state separate from governed workspace/runtime artifacts while preserving the existing relative runtime contract. Explicit ArtifactRoot overrides still work when operators need a different artifact location.
| Single Public Entry | |
|---|---|
| Install | β‘ Prompt-based install (recommended) |
| Public versions inside the same entry | Full Version + Customizable Governance / Framework Only + Customizable Governance |
| Result | choose host + action + version in one place, then copy the matching prompt |
The install surface is now registry-driven. HostId / --host selects host semantics, and the same public entry can resolve into governed, preview-guidance, or runtime-core depending on the adapter. full and minimal remain different products, but the difference is now mostly internal capability breadth and optional compatibility payloads, not "a few visible skills" versus "hundreds of visible skills." If you are not sure which path matches your host, start with the cold-start host matrix or the multi-host command reference.
When a follow-up step remains manual, the install docs now name the real path and file explicitly. For example: Codex uses ~/.codex/settings.json, Claude Code uses ~/.claude/settings.json, Cursor uses ~/.cursor/settings.json, OpenCode keeps the real ~/.config/opencode/opencode.json host-managed, and runtime-core hosts such as Windsurf / OpenClaw use <target-root>/.vibeskills/host-settings.json only as repo-owned sidecar state rather than an invented global settings contract.
The current install topology has been probe-validated against the installed runtime after each host-specific install / closure surface is materialized, not only against repo-local scripts:
| Host | Installed entry kept public | Installed-runtime probes covered |
|---|---|---|
codex |
skills/vibe public entry kept installed |
planning, debug, governed execution, memory continuity |
claude-code |
managed closure + installed skills/vibe payload |
planning, debug, governed execution, memory continuity |
openclaw |
runtime-core adapter + installed skills/vibe payload |
planning, debug, governed execution, memory continuity |
opencode |
preview-guidance adapter + installed skills/vibe payload |
planning, debug, governed execution, memory continuity |
Those probes validate that installed vibe still owns routing authority, preserves governance stage outputs, records cleanup receipts, and keeps memory read/write continuity after installation. They do not claim that every host-native public invocation syntax was exercised directly in the probe itself; the probe target is the installed runtime after the host-specific surface has been materialized. In a few planning-heavy scenarios, advisory gates can surface a bounded completed_with_failures status while runtime authority and governed delivery remain intact; that status is treated as expected only for explicitly allowlisted advisory checks.
β Custom workflow & skill onboarding guide
These capabilities weren't built from scratch. VibeSkills' foundation is the continuous integration of the best open-source solutions into one governed system.
We know that building in isolation can't keep pace with the rapidly evolving AI landscape. The core strength of VibeSkills comes from continuously absorbing the most mature methods and architectures from the open-source community, and bringing them under a single unified governance and orchestration system.
π Special Thanks & Acknowledgements
This project continuously integrates, absorbs, and governs the core strengths of the following excellent open-source projects:
superpowerΒ·claude-scientific-skillsΒ·get-shit-doneΒ·aios-coreΒ·OpenSpecΒ·ralph-claude-codeΒ·SuperClaude_FrameworkΒ·spec-kitΒ·Agent-SΒ·mem0Β·scraplingΒ·claude-flowΒ·serenaΒ·everything-claude-codeΒ·DeepAgentand moreThank you to all authors for your generous contributions β without these brilliant stars, VibeSkills would not exist. We have done our best to properly attribute and credit all absorbed repositories. If anything was missed, please open an Issue and we will correct it promptly.
You know what this is now. All it takes from here is one prompt:
β οΈ Invocation note: This project uses a Skills format architecture. Please invoke it through your host environment's Skills invocation method β do not run it as a standalone CLI program.
| Host Environment | Invocation | Example |
|---|---|---|
| Claude Code | /vibe |
I want you to design a XXX /vibe |
| Codex | $vibe |
I want you to design a XXXX $vibe |
| OpenCode | /vibe |
Use the vibe skill to plan this change. |
| OpenClaw | Skills entry | Refer to the host docs |
| Cursor / Windsurf | Skills entry | Refer to each platform's Skills docs |
π‘ Tip: To keep every message within the VibeSkills governed workflow, append
$vibeor/vibeto each of your messages. A message without the invocation syntax is treated as a regular request outside the governed runtime.
MCP note:
$vibeor/vibeis the governed runtime entry only. It is not MCP completion, and neither templates, manifests, sidecars, nor PATH-visible commands should be treated as proof that MCP is installed into a host's native MCP surface.
Currently supported public host surface: codex (strongest governed lane) Β· claude-code (supported install-and-use path with bounded managed closure) Β· cursor (preview-guidance path) Β· windsurf (runtime-core path with sidecar host-adapter state) Β· openclaw (preview runtime-core adapter path) Β· opencode (preview-guidance adapter path; direct install/check remains the thinner public path)
π Documentation & Installation Guides (click to expand)
Understand the system
Installation & Configuration
Give it a try! If you have questions, ideas, or suggestions, feel free to open an issue β I'll take every piece of feedback seriously and make improvements.
This project is fully open source. All contributions are welcome!
Whether it's fixing bugs, improving performance, adding features, or improving documentation β every PR is deeply appreciated.
Fork β Modify β Pull Request β Merge β
β If this project helps you, a Star is the greatest support you can give! Your support is the enriched uranium that fuels this nuclear-powered donkey π«
Thank you to the LinuxDo community for your support!
Tech discussions, AI frontiers, AI experience sharing β all at Linuxdo!