MIT · agent skills
◈
You are in The Armory — What will you teach it?
Skills your agent can actually run
Operational glue extracted from a real fleet: recovery routines, deploy helpers and secret-safety checks — packaged as skills an agent can load on demand.
# clone the pack git clone https://github.com/Aitherium/awskills.git # skills are plain markdown — point your agent at them ls awskills/skills/
87skills
MITlicense
0dependencies
Anyagent
Read the pack before you run it
Every skill in the distribution, with its own one-line summary. These are procedures written from incidents on a real fleet — open one and read it before you let an agent act on it in your environment.
skills/87 of 87
acp-drivedrive any external ACP agent from your agentacp-driveacp-serveexpose an AitherOS agent to ACP editorsacp-serveadk-deepseek-coderfill-in-the-middle, repo packing, and a reward that can failadk-deepseek-coderadk-harnessesone shell that drives every coding shelladk-harnessesadk-unslothlocal training and serving, wired to your agentadk-unslothadversarial-verificationstop grading your own homeworkadversarial-verificationagent-disk-hygieneyour coding agents are quietly eating your diskagent-disk-hygieneagent-integrationsaither talks to everything, everything talks to aitheragent-integrationsAgent operating doctrineagent-operating-doctrineagents-everywhereidea to a tool another agent can actually useagents-everywhereaither-agent-notebookturn "build X" into a reviewable, re-runnable planaither-agent-notebookaither-code-intelligencestand up agent code-search, and prove it actually worksaither-code-intelligenceaither-codegrapha call-graph-aware code index your agents can queryaither-codegraphaither-discord-agentdeploy your awdk agent as a Discord bot (automated onboarding)aither-discord-agentaither-headroomcut agent token cost with reversible context compressionaither-headroomaither-powered-codingmake any coding agent use AitherOS as its brainaither-powered-codingaither-prospectora semantic file-explorer that tells agents WHERE to lookaither-prospectorAither Retrievalone surface, every modalityaither-retrievalaither-startyou have a computer and an AI agent; let's make it yoursaither-startawconnectwire your machine, agent, and browser into AitherOSaitherconnectaithermeshstand up a private mesh and onboard nodes + agents to itaithermeshawnodeturn a machine's hardware into something your agents can useaithernodeaitherzeroprovision any machine from one config fileaitherzeroawconnectwire your machine, agent, and browser into AitherOSawconnectawdkrun your own AI agent, on your machine, in three commandsawdkawembedtrain an embedding model that knows your corpusawembedon your sidewhat the peer is missingawgit-claude-codeor from source:awgit-setupAither World Knowledgean operating doctrine you can installawknowledgebonsai-27brun a 27B model in ~3.8 GB on a plain CPU boxbonsai-27bbonsai2-27ba 27B reasoning model in 6–7 GB, on your GPU or your CPUbonsai2-27bCharacter Forgeconsistent characters, animation, talking avatarscharacter-forgeclaude-model-switcherkeep coding when Claude hits limitsclaude-model-switcherRenamed to `awknowledge`code-like-davidCompare two commitscompare-versionsConcurrent-safe gitconcurrent-safe-gitDebt Ledgerdebt-ledgerDecision Cardsstructured asks that actually reach the ownerdecision-cardsdeer-flowa long-running research/coding harness you host yourselfdeer-flowCheck outdated packagesdependenciesdgx-spark-deepseek-v4a 284B model at ~30 tok/s on one desk machinedgx-spark-deepseek-v4Do My Taxespersonal tax preparation workflowdo-my-taxesdocker-network-opsdiagnose container DNS and networking without fooling yourselfdocker-network-opsdocker-wsl2-build-safetystop bulk builds from killing Docker Desktop (and don't blame the disk)docker-wsl2-build-safetydocker-wsl2-disk-reclaimyour drive is full and `docker system df` is lying to youdocker-wsl2-disk-reclaimevent-loop-debuggerevent-loop-debuggerFail-Closed Authz Reviewfail-closed-authzfinetune-that-improvestrain your own model without making it worsefinetune-that-improvesFlociany cloud, locallyfloci-local-cloudGitHub Actions image pipelinegithub-actions-image-pipelineGitHub as an agentic control planegithub-agentic-c2github-artifact-quotathe storage limit that silently freezes your deploysgithub-artifact-quotaGitHub runner fleetone host, many runnersgithub-runner-fleetgpu-vram-attributionwhen the platform cannot tell you who is using the GPUgpu-vram-attributionor directly: adk ingest ./docs --agent research --chunk-size 1500 --chunk-overlap 200graph-a-repograph-rag-agentbuild a knowledge graph and an agent that owns itgraph-rag-agenthermes-agentNous Research's self-improving agent, on your hardwarehermes-agenthetzner-fleetprovision bare metal and VPS directly into your mesh from Hetzner Cloudhetzner-fleetinstall-skillsput these skills inside your agentinstall-skillslocal-image-generationthe agent draws on your own hardwarelocal-image-generationlocal-inferencea model on your box, for $0local-inferencemoat-guardmoat-guardModel fitnessbenchmarking a model for agent workmodel-fitness-benchmarkmodel-quantizationmodel-quantizationnvidia-cuda-mcpgive your agent current CUDA knowledgenvidia-cuda-mcpodsturn a box you already own into a private AI serverodsomninode-nodejoin the OmniNode P2P inference mesh in one commandomninode-nodeopenclawa personal assistant that runs on your machineopenclawProfile CPU usageperformancepooled-inferencerun a model no single box can hold, across every tier you ownpooled-inferenceProve It Liveprove-it-livePureCSS art → live HTML/CSS + PNG previewpurecss-artRamble-Driven Developmentramble-driven-developmentrecover-dockerrecover-dockerBeforerefactorRepo intakefrom "found a repo" to a decided outcomerepo-intakerepo-is-not-a-runtimethe architecture doctrine for agent-driven reposrepo-is-not-a-runtimerepo-to-websiterepo-to-websiteresume-allresume-allSafe bulk renamesafe-bulk-renamesecretguardsecretguardVULNERABLEsecurity-auditship-an-app-freeidea → public URL, $0, no server to rentship-an-app-freesplit-inferencesplit-inferencetaua minimalist terminal coding agent you can actually readtauwebsite-as-codewebsite-as-codeXY Chartsdata → picturexy-charts
What is in the pack
Each skill is a single markdown file describing a procedure precisely enough for an agent to execute it.
format
Plain markdown
No runtime, no framework, no build. A skill is a file — readable by a human and loadable by any agent that can read text.
recovery
Recovery routines
Procedures for getting wedged infrastructure back, written from incidents rather than from imagination.
deploy
Deploy helpers
The deploy-time checks and sequences that keep a change from being declared live before it actually is.
safety
Secret safety
Guardrails for handling credentials — what never goes in a commit, a log, or a tool argument.
portable
Agent-agnostic
Nothing here binds to one vendor's skill format; the content is the product.
free
MIT licensed
Free to use, fork and adapt, including commercially.
Quick start
Clone the pack and point your agent's skill loader at the skills directory.
git clone https://github.com/Aitherium/awskills.git cd awskills # browse what is available ls skills/ # for Claude Code, copy into your skills directory cp skills/*.md ~/.claude/skills/
These are procedures, not scripts. Read a skill before letting an agent act on it in your environment — they encode how this fleet is operated, and your topology may differ.
One estate, 8 doors
These projects are organs of the same system — the same identity plane, the same agent runtime, the same operational discipline. Follow any of them.
the osaitherium.comThe Living OS these four repos are cut from. Open, no login.pythonawdkBuild AI agent fleets — 3 lines, any backend, local or cloud.markdownawskillsPortable agent skills — self-contained procedures an agent loads on demand.markdownawknowledgeHow to run a coding agent so the result survives — the laws, with evidence.cudaaitherkvcacheNear-optimal KV cache quantization for LLM inference — sub-byte compression.powershellAitherZeroPowerShell 7+ automation framework — numbered, self-describing scripts.writingThe BlogHow all of it is actually built, measured and operated.pythonawrunA priority-aware queue and dispatcher for agentic runs and ad-hoc CI builds. It also judges whether the runner pool is big enough for the queue it is draining, and can ask a host to grow it -- reserving capacity is zero-sum, so a saturated pool needs more of it, not a different share of it.
The ecosystem this belongs to
Every Aitherium repository publishes an aither-manifest.json beside its
page, and each surface reads all of them — so the network is browsable from any node in
it. A repository that publishes nothing shows as unknown, never as zero.
The Aither World is an operating system for agents — a Linux you can hand to one, the runtimes it works in, and the tools it works with. awskills is one of its 66 bricks — each installs on its own, runs offline, and needs no account. All 66 →