Jannchie

Draft or rewrite GitHub Pull Request titles and descriptions — bug fixes, security fixes, features, refactors, dependency upgrades, rollbacks, plus release review notes and reviewer-facing change summaries. Use when the user mentions "写 PR", "PR 描述", "PR body", "review notes".

Write git commit messages that follow Conventional Commits. Keep them lean: the subject line states the intent in one line; add a body only when the diff can't speak for itself. Use when creating git commits, writing commit messages, splitting a working tree into commits, or when the user mentions "写 commit", "commit message", "提交信息", "提交规范".

Write or rewrite a project README that converts a 10-second visitor into a user. Use when the user asks to write/rewrite/improve a README, mentions "写 README", "README 写作", "项目说明", or wants a README generated from existing code or notes. Front-loads identity → value → minimum-runnable example; cuts marketing fluff, premature API dumps, and decorative TOCs.

Use this skill when setting up, validating, or debugging npm package release pipelines for pnpm workspace monorepos, especially with GitHub Actions, npm Trusted Publisher, provenance, scoped public packages, pack/publish tarball flows, and CI failures around workspace dependencies, Vue/TypeScript monorepo wiring, or npm release metadata.

Query and analyze machine-learning experiments tracked by pandm — list runs, read a run's config/summary/metrics, compare runs to find the best hyperparameters, read the display specs the training code declared (which metric is the experiment's judge, which invariants must hold), read full metric series, and locate logged images. Reads a local `.pandm/` SQLite + PNG store directly; fully offline, no network, account, or server required. Use when the user asks about past pandm experiments, wants to compare runs, pick a winner, check whether a run broke an invariant, inspect a metric over time, or analyze results stored in a `.pandm/` directory.

Record machine-learning experiment metrics, images, distributions, hyperparameters, and training progress to pandm — a local-first, offline-by-default, account-free wandb/tensorboard alternative that writes to a local `.pandm/` SQLite + PNG store. Needs no network, server, login, or API key; a shared server is strictly optional, and once one is configured the same code auto-pushes to it in the background. Use whenever you write OR edit pandm instrumentation: instrumenting a training/eval loop, logging scalar metrics or images, adding/renaming/regrouping a logged metric, changing what a script reports, shaping how a chart renders (titles, subtitles, units, axis labels, panels, confidence bands, bar/histogram charts — especially for RL), reporting an ETA, or saving a run's config to compare in the dashboard. Read this before touching any `pandm.init` / `run.log` / `run.define_metric` call — it sets the titling discipline every run and every metric needs.

Query and analyze machine-learning experiments tracked by pandm — list runs, read a run's config/summary/metrics, compare runs to find the best hyperparameters, read full metric series, and locate logged images. Use when the user asks about past pandm experiments, wants to compare runs, pick a winner, inspect a metric over time, or analyze results stored in a `.pandm/` directory.

Record machine-learning experiment metrics, images, hyperparameters, and training progress to pandm — a local-first, account-free wandb/tensorboard alternative that writes to a `.pandm/` SQLite + PNG store. Use when instrumenting a training or evaluation loop, logging scalar metrics or images from a Python script, reporting an ETA, or saving a run's config so it can later be compared in the pandm dashboard.

Use arthash to add tiny placeholder-image hashes (17–400 bytes) to an application — the artistic alternative to blurhash / thumbhash / sqip. Trigger this skill when the user wants to render a low-quality image preview (LQIP) while a real image loads, replace blurhash or thumbhash, generate SVG placeholders without spawning a Go subprocess (sqip), or produce a stylised mosaic from a palette. arthash ships three SDKs (Rust crate, Python `arthash`, npm `arthash`) that all read and write the same byte format, so a hash made on the server decodes identically in the browser. This skill covers picking the right SDK, the right mode (default `RECT` and other shape modes `TRIANGLE` / `CIRCLE` / `SQUARE` / `ROTATED_RECT` for SVG, `PIXEL` for mosaic, `DCT` only when you specifically want a blurhash/thumbhash-style blurry look), the right palette, the encode → store → decode → render lifecycle, and the small pitfalls (codecs are not in the bytes; inputs must be small; SVG modes vs raster modes). Per-language API details l

Browse snowcake47's illustration portfolio (cake47.art) — fan works, original characters, and commercial commission samples. Use when a user asks about snowcake47 / 私期 / cake47, requests anime-style illustration samples by this artist, or wants commission contact channels.

Use cairnq, the cross-language durable task runtime that coordinates through a shared database (SQLite for one host, Postgres for many). Covers the worker side (register handlers, run them, batch delivery), the API side (submit / call / get / cancel / retry / stats by id or business key), idempotency keys, retries, cooperative cancel, and the at-least-once limits. Trigger when code imports `cairnq` (Python or TypeScript), defines a Worker or handler, calls submit/call/getByKey, batches tasks into one handler call, or the user mentions cairnq, tasks.db, a durable task queue, or an embedded SQLite job runtime.