Latest Agent Skills
Everything added to the registry recently, newest first.
- seismoforg BundlePlan
Plan a change with the user through multiple-choice questions, store it in docs/plans/, then build it with parallel subagents, verify it and close it via the commit skill. Use when the user wants to plan, build or resume a feature, or runs /plan.
- fsantibanezleal SkillContext Audit
Use when a session feels heavy or before adding rules or memory: measures what loads at session start and checks it against the budgets.
- fsantibanezleal SkillDelegate
Use when launching a named agent on a linted brief, on the subagent lane, the Agent SDK lane or a provider lane, and when accepting the agent's report.
- fsantibanezleal SkillPromote
Use when every gate through L9-promote is green and a human wants to merge a unit: merges only on green checks for the exact head commit.
- fsantibanezleal SkillLessons
Use when a unit's build has passed, to record its dated lesson (L10-lessons) and check LESSONS.md and overdue ADR reviews.
- fsantibanezleal SkillVerify
Use at the end of every gate: an independent agent with fresh context reproduces each claim at the reviewed commit and writes the evidence file.
- fsantibanezleal SkillFreeze
Use when the spec, contracts and ADRs are ready and a human is about to approve them (L5-freeze). Prints the phrase to type and checks that the approval still matches.
- fsantibanezleal SkillDoctor
Use when Batuta seems not to run, after an install or update, or on a new device: checks the setup, routing, contracts, hooks wiring and machine limits.
- fsantibanezleal SkillRoute
Use when choosing which agent and model runs a task, or before launching any agent. Prints the role table (model, effort) from batuta.json and checks the agent files against it.
- fsantibanezleal SkillGreen
Use when locked red tests exist and the code must make them pass (L7-build, green): minimal code, a [green] commit, history check and trace.
- fsantibanezleal SkillBrief
Use when about to hand work to an agent: write the brief first. Creates a brief from the template with its ledger row and lints it before launch.
- fsantibanezleal SkillSpec
Use when starting a feature or product, before any code: understand, size, decide technology, write contracts and the spec (L0 to L5, up to the freeze).
- fsantibanezleal SkillGate
Use when asking whether a lifecycle gate is passed or what blocks the next one: status, evidence files and the deterministic checks for L0 to L10.
- fsantibanezleal SkillRed
Use when a task's tests must exist before its code (L7-build, red): a test-author writes failing tests, they are locked by hash and a [red] commit records them.
- taskade BundleTasks Md
Create and maintain a TASKS.md build ledger at the repo root so work survives a context reset. Use when starting a multi-step task, resuming an interrupted one, when the user asks to track progress or "remember where we left off", or when a request carries several clauses that can be silently half-built. Not for a to-do queue of future work or a one-off question that changes no files. Defines six sections (Destination, Open decisions, Out of scope, Requirements, Done, Knowledge), the Requirements block with its States line, the dated-entry format, and the evidence bar a clause must clear before it is recorded as done.
- mcf-studio BundleGodot Headless Test
Lancer et lire des tests Godot 4 en headless sans son et sans toucher aux sauvegardes ni aux réglages de l'utilisateur (données user:// isolées), avec une sortie réduite aux résultats. À utiliser pour tester une modification d'un projet Godot, relancer un test existant ou écrire un nouveau test automatisé.
- mcf-studio SkillGodot Editor Scenes
Règles et pièges pour créer ou modifier des scènes (.tscn), nœuds et scripts d'un projet Godot 4 pendant que l'éditeur est ouvert, à la main ou via un serveur MCP Godot. À utiliser avant de modifier une scène ou un script d'un projet Godot dont l'éditeur peut être ouvert, et avant de commiter une scène réenregistrée par l'éditeur.
- mcf-studio SkillEconomie Tokens
Conseils pour consommer moins de tokens avec Claude Code - quand faire /clear ou /compact, quel modèle choisir, comment formuler une demande, alléger CLAUDE.md, les MCP et les sous-agents. À utiliser quand l'utilisateur dit qu'il atteint sa limite trop vite, demande pourquoi ça consomme autant, ou veut optimiser sa configuration.
- 4miers BundleRelay
Hand off to a fresh agent session at the end of a task instead of compacting mid-work. Use when the user asks to hand off, wrap up for a new session, save the session state before clearing the context, or when they ask to resume, pick up the handoff, or continue where the last session left off.
- 4miers SkillHandoff
Write a handoff of the current work so a fresh session (after /clear) can pick it up. Use when the user asks to hand off, wrap up for a new session, or save the session state before /clear.
- 4miers SkillResume
Continue from a relay handoff in this project (.claude/relay/handoffs/ or .claude/relay/done/). Use when the user asks to resume, pick up the handoff, or continue where the last session left off.
- anderson-gomes-lab SkillMethod
Run a project with a flight plan shown in the Pilotage pane. Use it when the person starts or resumes any project with a deliverable and several steps (website, app, automation, content), when a PILOTAGE.md exists in the project, when a second workstream appears, and when they say "where are we", "I'm lost" or "how many agents". Stops to write PILOTAGE.md first, routes each message in one line, keeps "À toi" true, closes each batch.
- qubehub-ai SkillInventory Lookup
Answer questions about a developer's live inventory in QubeHub — available units, prices, statuses, unit stats per project, price history — through the QubeHub connector. Use for "what is available in", "price of unit", "how many 2-beds are left", "did the price change", unit selection for a buyer.
- qubehub-ai SkillSales Playbooks
Apply the QubeHub sales playbooks (lead qualification, follow-up cadence, objection handling, closing and reservation, selection presentation) to a concrete lead or situation, with the workspace's own data from the QubeHub connector. Use for "how do I handle this objection", "what should I ask this lead", "write a follow-up", "how do I move this to reservation".
- qubehub-ai SkillPipeline Review
Review a sales pipeline in QubeHub — stuck leads, overdue tasks, deals without next steps, stage conversion — through the QubeHub connector and produce a prioritized action list. Use for "review my pipeline", "what should I focus on today", "which deals are stuck", "weekly sales review", head-of-sales check-ins.
- qubehub-ai SkillLead Briefing
Prepare a sales manager for a call or meeting with a QubeHub lead — pull the lead, its contact, deals, tasks and recent activity through the QubeHub connector and turn them into a one-screen brief. Use when asked to "brief me on", "prepare me for the call with", "what do I need to know about" a lead, buyer or contact.
- satisfying-systems BundleHumanize Without Changing Facts
Use when AI-written prose (articles, emails, docs, captions) needs to read like a person wrote it while its facts stay put. Runs an editor pass over prose only, then rejects edits that change per-citation numbers, double-quoted spans, links, code, citation placement or structure, or that a separate judge model says changed the meaning.
- satisfying-systems BundleWriter Reviewer Split
Use when you hand coding work to an AI agent and want it checked before it ships. One model writes the change in an isolated worktree, a different model reviews the exact commit read-only, and nothing merges unless the reviewer's final line is VERDICT: PASS.
- matchalatteguy SkillFaintmark
Forensic (per-recipient) invisible image watermarking. Create one marked copy of an image per recipient, and later identify which registered copy a shared or leaked image is. Use when the user wants to share a picture with several people through chat apps (WhatsApp, Discord, Telegram, iMessage, Signal, WeChat, X) or email and trace a copy back to its recipient, or asks which recipient a received image was issued to. Not for visible logos or text watermarks.
- jsmak BundleLecture Slides
Build a Korean in-house lecture (딥러닝 강의 style) from the lecturer's drafts: a Slides artifact deck with speaker notes, a polished final script, an ElevenLabs phonetic script (소리 나는 대로 + break tags), and — when the lecture has a practice notebook — a renamed, cleaned .ipynb pushed to the course GitHub repo with a Colab badge and README row. Use when the user hands over "대본 초안", "PPT 초안", "슬라이드 초안" or an .ipynb for the next lecture ("N강을 만들려고 합니다", "슬라이드와 대본을 만들어 주세요", "최종 대본과 일레븐랩스 대본", "슬라이드를 이전 스타일과 유지되도록"), or asks to revise a lecture deck, sync notes with a reviewed script, or add figures/formula images to a lecture slide. The narrated video itself is the lecture-video skill's job.
- alchemy-lab-infra SkillAI Regulation
The current state of AI regulation and AI legislation — the EU AI Act, US federal bills and agency actions, US state AI laws, and UK AI policy — with each item's legislative stage. Use when the user asks where a law stands, what AI regulation passed or is proposed, "AI Act status", "AI regulation tracker", or what an AI bill does. Live, sourced from Congress, state legislatures, UK Parliament and EU bodies; free and no key.
- alchemy-lab-infra SkillAI Receipts
How much of what is being said about an AI company, model, or lab is self-reported versus independently evaluated or reported — the receipts behind the claims. Use when the user asks whether a benchmark or capability claim is verified, "is this independently confirmed", who is actually reporting something, or how to weigh a lab's announcement. Also for what regional press covers that Western feeds miss. Free and no key.
- alchemy-lab-infra SkillAI News
What happened in AI today, this week, or on a given date. Use when the user asks for AI news, an AI briefing, "what's new in AI", what a lab or model announced, or what is trending in AI — and for follow-ups on a specific topic, region, or company. Backed by Horizon's daily briefing (Hacker News, Bluesky, lab blogs, arXiv, Hugging Face, regulators), free and no key.
- skyf0xx SkillDistribution
Use on a business goal when the question is how customers will find and buy it — no sales yet, sales that come only through friends, a launch with no plan for reach, or effort spread across many channels with nothing working. Checks for product-market fit with Sean Ellis's survey, then runs Weinberg & Mares's Bullseye to test a few channels cheaply and concentrate on the one that works.
48 - skyf0xx SkillPositioning
Use on a business goal when it's unclear who the product is for, what it should be compared with, or why anyone picks it — before pricing, a launch, a pitch or a website rewrite, or when people who see it don't get it. Works April Dunford's positioning sequence from the customers who already love it to a committed positioning, recorded as a decision with a reverse-if.
48 - skyf0xx BundleOrganizing
Use on an activism goal to build the people power a campaign needs — when the same few activists do everything, turnout is high once and then fades, the movement has no structure or spokespeople, or the campaign is about to escalate without knowing whether the majority will follow. Separates organizing from mobilizing, charts the whole constituency, finds the organic leaders, prepares the one-to-one organizing conversation, and sets structure tests with thresholds fixed in advance.
48