AI & ML

mzvonar Bundle

Updates the repo's canonical guideline docs (CLAUDE.md, AGENTS.md, architecture docs) and propagates the change to every project agent and skill that generates code. Use when the user says "update the guidelines", "update CLAUDE.md / AGENTS.md", "update the architecture", "incorporate this research"; when a convention changes systematically (naming, layering, validation, dependency, test or commit rules); when a lesson promoted by /workflow:lessons needs a guideline gist plus a skill body; or when the user asks to make CLAUDE.md leaner / slimmer / shorter, do "CLAUDE.md hygiene", or trim / de-bloat it (then the hygiene runbook is the whole job). Invoke proactively on any systematic change to how code should be written. Always runs every phase (read, update docs, hygiene, analyse managed files, apply) and never stops after one doc: a stale assertion is worse than a missing one. CLAUDE.md stays an index of one-line gists under 400 lines / 40,000 chars; rule bodies live in the repo's project skills.

mzvonar Skill

Prepare a clean-context handoff. Use when the user asks to "hand off", "prepare a clear-context handoff", "write a handoff", or otherwise wants to /clear and continue in a fresh session. Produces a committed handoff doc (what we're doing, what's done, ordered next steps, research/open questions) plus a kickoff prompt to paste into the fresh context.

stifleur390 Bundle

Créer un nouveau skill agent multi-plateforme. Utiliser quand l'utilisateur veut créer un skill, ajouter un skill au repository, transformer des instructions répétitives en skill, ou documenter un workflow réutilisable. Couvre la rédaction du SKILL.md, le choix de la structure, la rédaction de la description, le déploiement multi-agents, et la publication. Ne PAS utiliser pour modifier un skill existant (éditer directement le SKILL.md).

kroshank Skill

Break a complex goal into a step-by-step workflow, assign tools to each step, and optimize execution. Use when the user wants an SOP, agent workflow, automation plan, or multi-step execution design.

provectus Bundle

This skill should be used when the user asks to "create an MCP server", "build an MCP tool", "add an MCP resource", "define MCP prompts", "set up FastMCP", "run an MCP server", "expose tools via MCP", "mount MCP sub-servers", "configure MCP transport", "add authentication to MCP", "test an MCP server", "use the MCP client", or when writing any Python code that uses the fastmcp package. Provides up-to-date FastMCP API patterns for tools, resources, prompts, server composition, authentication, and deployment.

provectus Bundle

This skill should be used when the user asks to "set up ChromaDB", "create a Chroma collection", "add embeddings to ChromaDB", "query ChromaDB", "search vectors", "semantic search with ChromaDB", "filter ChromaDB results", "ChromaDB metadata filtering", "configure Chroma", "use ChromaDB persistent client", "delete from ChromaDB", or when writing any code that interacts with the chromadb Python package. Provides up-to-date API patterns, filtering syntax, collection configuration, and embedding function integration.

g-tavares14 Skill

Delivers changes incrementally in thin, verifiable slices. Use when the user runs /agent-skills:build. Use when implementing any feature or change that touches more than one file, or when picking up the next task from a plan. Use when rolling a change out behind a feature flag, when you're about to write a large amount of code at once, or when a task feels too big to land in one step.

g-tavares14 Skill

Breaks work into ordered tasks. Use when the user runs /agent-skills:plan. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.

808enzo Bundle

Write down what a metric actually is, so two people quoting it mean the same thing. Use when two dashboards disagree, when a metric is about to become a target, when you need the numerator, denominator, window and attribution method of a lifecycle metric written so someone else can reproduce it, when you are reconciling a number across two systems, or when you are deciding which metrics a program reports on and who owns them. Covers message and channel metrics, customer and money metrics, the definition-change protocol, and the self-baseline you build when no citable benchmark exists. Not the dashboard, not the attribution model, not the proof that a difference is real.

drdustinedwards Skill

Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.

drdustinedwards Bundle

Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).

drdustinedwards Skill

Rewrites model-facing lines in this repo's skills and commands to match a named model's prompting guide. Use when a guide in guides/ has changed, when a new model appears, or when asked to refresh the skills for a model. Rewrites calibration only; it never edits a gate, a ruling, or task content.

808enzo Bundle

Take a loyalty program from a signed design to a running one, and get people into it. Use when deciding whether the program is worth launching at all, checking that the design can actually run at the till, sequencing the build, sizing a pilot, rolling out by location or channel, migrating members off an old program, removing friction from enrollment, getting store staff to ask, or reading the first months before any effect is readable. Not the earn-and-burn model, not reward value or tiers, not discount depth, and not the day-to-day operation of a program that is already live.

808enzo Bundle

Design a comparison a decision can rest on. Use when writing a testable hypothesis, choosing what to randomize, sizing a test and deciding how long to run it, building a local or global control group, sizing a holdout and pricing what it costs, checking whether a running test is still valid, reading a result without over-reading it, or answering whether a mechanic produces anything at all beyond what customers would have done on their own. Covers A/B design, holdouts drawn from flow entries, verifying an exclusion, incremental measurement of an added cascade step, the evidence ladder, and the test log. Not the metric formula, not the dashboard, not the attribution model.

jtydhr88 Bundle

Synth and timbre design as an arrangement decision (音色设计) - deciding what a sound must be before touching a knob. The four-question intent checklist (percussive or pitched / register / envelope shape / bright or dark) and its mapping to oscillator, tune, amp EG and filter; resetting to an unprocessed starting state instead of browsing presets; the sculpting model of subtractive synthesis; filter selection; and why a pure sine bass vanishes on small speakers (difference tones). Use when choosing or designing a patch, when a synth sound is generic or preset-sounding, when a bass disappears on phones or laptops, when a pad has no character, or when writing the timbre descriptors of an ARR-SPEC roster. 音色设计、合成器、滤波器、波封、预设、贝斯消失、差音。

drdustinedwards Skill

Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria.

qf-studio Bundle

Web deep research producing a cited report from fetched sources, adversarially reviewed and gate-checked, with conclusions ingested into the knowledge graph. Auto-invoke when user says "deep research on", "research the web for", "write a research report on", "what does the literature say about", or "deep dive into" a topic outside the codebase. For codebase questions use the navigator-research agent instead.

jtydhr88 Bundle

Compiling ARR-SPEC into symbolic notation (符号乐谱输出) - ABC, MusicXML and MIDI from one emitter. Why the symbolic path matters (it is the only fully reproducible output and the only way YuE2 executes rather than guesses), the field-by-field mapping from spec to ABC headers and body, roman numerals to letter chords, multi-voice and per-part output, what each target format can and cannot carry, MIDI CC and velocity for expression, and the verification loop that catches a bad emit before it reaches a synth. Use when producing ABC for YuE2, when exporting MIDI or MusicXML for a DAW, when a deterministic reference render is needed for A/B, or when an LLM should write notation directly. 符号乐谱、ABC、MusicXML、MIDI、打ち込み、直出乐谱、确定性渲染。

g-tavares14 Skill

Simplifies code for clarity. Use when the user runs /agent-skills:code-simplify. Use when refactoring code for clarity without changing behavior. Use when code works but is harder to read, maintain, or extend than it should be. Use when reviewing code that has accumulated unnecessary complexity.

understudylabs Bundle

Use when a developer already has production LLM traces — a bucket of captures, provider log exports, or gateway capture files — and wants them visualized, turned into local redacted eval sets, or profiled for cost. "Visualize this trace", "ingest my traces", "turn these logs into an eval set", "where is my LLM spend going".

understudylabs Skill

Compatibility shim for Claude Code installs. Use when a developer asks to install, update, enable, reinstall, remove, or verify the Understudy Claude Code plugin - "install Understudy in Claude", "add the Understudy skills", "why can't Claude see the skill". Route to install-agent-adapter with platform claude-code.

understudylabs Bundle

Use after a route decision exists and a candidate model must take live traffic safely — "ramp this route", "move 25% of traffic to the new model", "did the route change regress anything", "roll this back", "prove the savings are real". Pre-ramp stability gates, a staged traffic ladder through the Understudy gateway dial, routed-vs-passthrough verification from captures at each step, and explicit rollback triggers.

understudylabs Bundle

Use when a developer wants to build an eval from their real LLM app before changing anything — "measure how my app is doing today", "build an eval from my workload", "we have no baseline", "is my current model actually good". Turns the workload into auditable local artifacts (harness, metric, frozen splits, baseline); has a public-benchmark on-ramp when no traces exist.

understudylabs Bundle

Use when a developer has a measured eval and wants the prompt or route improved without retraining — "optimize my prompt against my eval", "close the quality gap with the cheaper model", "squeeze cost without losing quality". Keeps holdout sealed, runs GEPA (automatic prompt evolution) on train/dev only, and requires a claim packet before any savings statement.

understudylabs Bundle

Use when a user asks what Understudy is, how Understudy Desktop works, how local model serving, Fusion sidekick, evals, model candidate results, rollout labs, Product Knowledge, or Understudy product capabilities should be explained to developers, customers, or agents.

understudylabs Bundle

Use when a developer wants to replace an expensive frontier model on a classification workload (binary, multi-class, multi-label, or structured extraction) with a fine-tuned open-weight student — "distill this classifier", "can a small model do this tagging job", "the frontier labels these for $X, make it cheaper", "consensus-label my data". Multi-teacher majority-vote labeling, failure-directed SFT data, and a four-way promote/shadow/collect/stop verdict.

understudylabs Skill

Use when a coding agent has edited a user's LLM app and wants a regression verdict on the frozen benchmark tasks — "did my code change regress the eval", "run my actual app against the benchmark", "author an app-harness.json". Drafts the understudy.app_harness.v1 sidecar from the user's repo, queues an app_replay run, and reads the honest (marked-anomaly, never fabricated) rows.

g-tavares14 Skill

Optimizes agent context setup. Use when starting a new session, when agent output quality degrades, when switching between tasks, or when you need to configure rules files and context for a project.

g-tavares14 Skill

Conducts multi-axis code review. Use when the user runs /agent-skills:review. Use before merging any change. Use when reviewing code written by yourself, another agent, or a human. Use when you need to assess code quality across multiple dimensions before it enters the main branch.

g-tavares14 Skill

Drives development with tests using the red-green-refactor loop. Use when the user runs /agent-skills:test. Use when implementing any logic, fixing any bug, or changing any behavior. Use when you need to prove that code works, when a bug report arrives, or when you're about to modify existing functionality.

understudylabs Bundle

Use when a developer wants to compare candidate models — any mix of local, gateway, or frontier — on the same eval and see quality, latency, cost, and reliability side by side. "Which model should I use", "sweep these models on my benchmark", "compare Gemma vs the frontier on my eval". To stand up and serve a local candidate first, use run-local-model-lab.

understudylabs Skill

Use to turn loose per-task trajectory JSONs (a Lilac export or local rollout corpus) into a queryable, provenance-tracked, contamination-safe dataset that hard-blocks any selection leaking frozen dev/holdout into a train/RL/distill pool. "curate my trajectories", "which rollouts are train-safe", "exclude the holdout rows before RL", "is this distill pool contaminated", "make a hash-stamped selection", "stop hand-filtering trajectories in bash".

understudylabs Bundle

Use when a developer wants to stand up and run a local model on Apple Silicon against their real workload — "run this model on my Mac", "is a local model good enough before I pay for hosted". Covers the MLX serving rig, scored real-workload evals, and the route decision. For comparing many candidate models on one eval, use compare-model-sweep.

understudylabs Bundle

Use when a developer wants a captured prompt, trace, or dataset explained before changing anything — "explain this prompt", "what is this trace actually doing", "decompose this workload", "help me understand this dataset before testing models". Builds a shared mental model (purpose, inputs, tools, success criteria) with the user through Q&A.

understudylabs Skill

Measure per-band model headroom from a local, source-bound run artifact before investing in training.

understudylabs Bundle

Use when you need to know HOW two model runs differ behaviorally on the same tasks, not just THAT one scores higher — per-task trajectory diffing that classifies the gap as persistence/recovery, knowledge, or format/parsing. "why does the bigger model pass these", "is this gap RL-shaped", "diff these two trajectory runs", "where do the trajectories diverge", "what would distillation buy me". The behavioral complement to compare-model-sweep.