New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Soup

by sorryhyun in sorryhyun/anima_lora

Run or modify the uncond-init soup pipeline (make soup, scripts/soup/, configs/soup/soup.toml) — selection globs, uncond-init reuse and naming, env knobs, sigmalowres interaction, and constraints.

MITAuto-check passed

Install Soup

skills CLI
$ npx skills add sorryhyun/anima_lora --skill soup -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install sorryhyun/anima_lora soup --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/sorryhyun/anima_lora.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/soup .claude/skills/soup && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
soup
GitHub stars
125
Token cost
~728 tokens
SKILL.md length
323 words
Files
1
Skills in repo
12
Repo updated
First seen
Licence
MIT

At a glance

Run or modify the uncond-init soup pipeline (make soup, scripts/soup/, configs/soup/soup.toml) — selection globs, uncond-init reuse and naming, env knobs, sigmalowres interaction, and constraints.

  • Works in 3 steps: Uncond inter-train — a short… → Fine-tunes — 3 seeded fine-tunes on… → Soup — exact ΔW average, SVD-truncated…
  • SKILL.md covers What it does, Selection, Knobs and Config, plus 1 more section
  • Calls make

What it does

Soup is an agent skill from sorryhyun/anima_lora. Run or modify the uncond-init soup pipeline (make soup, scripts/soup/, configs/soup/soup.toml) — selection globs, uncond-init reuse and naming, env knobs, sigmalowres interaction, and constraints. Load before running a soup, changing its knobs, or editing scripts/soup/.

Its SKILL.md is about 730 tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

The repository describes itself as: optimized anima lora training script. The licence is MIT.

Example prompts

  • “/soup”

Workflow steps

3 steps, taken from the first numbered list in SKILL.md.

  1. Uncond inter-train — a short caption_dropout_rate 1.0 train on a diluted pool: POOL_PATH_PATTERN (default * = whole dataset), diluted by…
  2. Fine-tunes — 3 seeded fine-tunes on PATH_PATTERN starting from that init.
  3. Soup — exact ΔW average, SVD-truncated back to network_dim → anima_soup_.safetensors + snapshot (slug derived from the pattern or NAME).

What it can do on your machine

Read from SKILL.md and the folder at commit d16b651. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • make

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Soup loads about 728 tokens when it runs. Until then it costs about 69 tokens; SKILL.md has 323 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~69
When it runs · the whole SKILL.md, loaded when a task matches
~728

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from sorryhyun/anima_lora at commit d16b651, republished under its MIT licence (© sorryhyun). 323 words, ~728 tokens.

Download SKILL.mdSave it as .claude/skills/soup/SKILL.md (or your agent's skills folder).
name
soup
description
Run or modify the uncond-init soup pipeline (make soup, scripts/soup/, configs/soup/soup.toml) — selection globs, uncond-init reuse and naming, env knobs, sigma_lowres interaction, and constraints. Load before running a soup, changing its knobs, or editing scripts/soup/.

Uncond-init soup pipeline

make soup runs the uncond-init soup pipeline (scripts/soup/; GUI: Experimental tab → soup). Recipe from bench/memorization/report.md; deep-dive: docs/experimental/soup.md.

What it does

Three stages, submitted as one daemon command job (attach-by-default; --queue detaches, --inline bypasses the daemon):

  1. Uncond inter-train — a short caption_dropout_rate 1.0 train on a diluted pool: POOL_PATH_PATTERN (default * = whole dataset), diluted by --sample_ratio, with the fine-tune set always unioned in. The checkpoint name is deterministic and the checkpoint is reused if it exists — one shared uncond init serves every soup drawing the same pool.
  2. Fine-tunes — 3 seeded fine-tunes on PATH_PATTERN starting from that init.
  3. Soup — exact ΔW average, SVD-truncated back to network_dim → anima_soup_<slug>.safetensors + snapshot (slug derived from the pattern or NAME).

Selection

make soup PATH_PATTERN="<glob>", or the TARGET=<dir> shorthand ⇒ PATH_PATTERN="<dir>/*" NAME=<dir>. Selection is an fnmatch path_pattern, not an artist dir (| = alternatives) — a soup can span any glob slice.

Knobs

env: NAME, POOL_PATH_PATTERN, UNCOND_RATIO, UNCOND_EPOCHS, UNCOND_INIT, NUM_SOUP, RANK, ARTISTS_SHARD, CUSTOM, plus LR_POOL/LR_INTERVAL for opt-in per-ingredient LR diversity. By default ingredients vary by seed only; the soup is a uniform average with no greedy-selection gate, so a bad LR draw is averaged in, not dropped.

NO_UNCOND=1 / [soup] no_uncond = true skips Phase 1 — ingredients init from down_init (weight_svd [+ top-level svd_slice]); required for svd_slice != 0 (the uncond --network_weights warm-start would overwrite the slice; the pipeline refuses). All ingredients of one soup share one slice; the slug gets _nouncond[_k<slice>].

ARGS reaches the fine-tunes — except --sigma_lowres*, a whole-pipeline routing knob that is also replayed onto the uncond run and folded into its name as an _sl<digest> tag, so a σ-demoted soup can't silently reuse a natively-trained init (and vice versa). Names are unchanged when σ-demote is off.

Config

Ingredients train --method soup → configs/soup/soup.toml, a dedicated plain-LoRA stack. Its [soup] table holds the pipeline knob defaults (incl. pool_path_pattern) and is stripped from the train.py merge.

Constraints

  • Plain-LoRA checkpoints only — Hydra (and any non-plain-LoRA key shape) refused.
  • Quality win + seed-lottery insurance, NOT a memorization fix.

© sorryhyun, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/soup of sorryhyun/anima_lora.

Open the folder on GitHubat commit d16b651

Compare with similar skills

Soup next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Soup compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Soup this skillsorryhyun/anima_lora125—~728Automated safety check: PassMIT
Initasgeirtj/system_prompts_leaks69k—~5.5kAutomated safety check: PassCC0-1.0
Initasgeirtj/system_prompts_leaks69k—~602Automated safety check: PassCC0-1.0
Initbrycewang-stanford/Auto-Empirical-Research-Skills4.6k—~596Automated safety check: NotesCustom licence
Agent Coordinator Swarm Initruvnet/ruflo74k2 repos~997Automated safety check: PassMIT
Swarm Initruvnet/ruflo74k—~319Automated safety check: PassMIT

Similar skills

  • Init

    asgeirtj/system_prompts_leaks

    Initialize new CLAUDE.md file(s) and optional skills/hooks with codebase documentation

    69k GitHub stars~5.5k tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Init

    asgeirtj/system_prompts_leaks

    Initialize a new CLAUDE.md file with codebase documentation. An agent skill from asgeirtj/system_prompts_leaks.

    69k GitHub stars~602 tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Init

    brycewang-stanford/Auto-Empirical-Research-Skills

    Fills all [FILL:] placeholders across the template to initialize a new research project.

    4.6k GitHub stars~596 tokensUpdated 4 days ago
    Auto-check: notes

  • Agent Coordinator Swarm Init

    ruvnet/ruflo

    Agent skill for coordinator-swarm-init - invoke with $agent-coordinator-swarm-init

    74k GitHub starsUsed in 2 repos~997 tokens
    Auto-check passed

  • Swarm Init

    ruvnet/ruflo

    Initialize a multi-agent swarm with anti-drift configuration.

    74k GitHub stars~319 tokensUpdated today
    SecurityAuto-check passed

  • Federation Init

    ruvnet/ruflo

    Initialize federation on this node — generate keypair and configure peers

    74k GitHub stars~221 tokensUpdated today
    Auto-check passed

More from sorryhyun/anima_lora

All 12 skills in this repo

  • Model Catalog

    sorryhyun/anima_lora

    The model catalog (library/downloads.py) — one Asset row per weight (repo, files, destination, installed probe), packs, resolve() name order, and the rule that loaders import their default paths…

    125 GitHub stars~502 tokensUpdated today
    Auto-check passed

  • Qwen21

    sorryhyun/anima_lora

    Qwen-Image-2.1 LoRA line (NOT Anima) — running cache/train through the daemon, make gui-qwen, the CacheRequest/TrainRequest flag surface and how to add a field, model-dir resolution, cache layout…

    125 GitHub stars~1.9k tokensUpdated today
    Auto-check: notes

  • Anime Tools

    sorryhyun/anima_lora

    The trainer ↔ animetools boundary — what the curation split moved out, the typed request/stage API the make targets build, the git-pin dev loop and its stale-venv trap, and the tests that guard the…

    125 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Bucketing

    sorryhyun/anima_lora

    Free-fit native-shape bucketing — the token bands per edge tier, tier choice at preprocess time, the compiledynamicseq coupling and per-tier graph budget, and why training never needs --targetres.

    125 GitHub stars~932 tokensUpdated today
    Auto-check passed

  • Captions

    sorryhyun/anima_lora

    Caption pipeline — position-clause grammar (never hand-split a caption), make caption-autotag modes, make caption-position (v2 rewrite rules and gates), and the preprocess-stage wiring for both.

    125 GitHub stars~1.3k tokensUpdated today
    Auto-check passed

  • Custom Nodes

    sorryhyun/anima_lora

    The ComfyUI node map — which node lives in which standalone repo vs in-tree under customnodes/, where each is symlinked, and the vendor-sync rule for the vendor/ subsets.

    125 GitHub stars~470 tokensUpdated today
    Auto-check passed

Questions about Soup

What does Soup do?

Run or modify the uncond-init soup pipeline (make soup, scripts/soup/, configs/soup/soup.toml) — selection globs, uncond-init reuse and naming, env knobs, sigmalowres interaction, and constraints. Soup is an agent skill from sorryhyun/anima_lora.toml) — selection globs, uncond-init reuse and naming, env knobs, sigmalowres interaction, and constraints.

How do I install Soup in Claude Code?

Run `npx skills add sorryhyun/anima_lora --skill soup -a claude-code`. Or copy the skill folder (.claude/skills/soup in sorryhyun/anima_lora) into .claude/skills/soup in your project. Claude Code loads it when a task matches its description.

How do I install Soup in Codex?

Run `npx skills add sorryhyun/anima_lora --skill soup -a codex`. Or copy the skill folder (.claude/skills/soup in sorryhyun/anima_lora) into .agents/skills/soup in your project. Codex loads it when a task matches its description.

Can I use Soup in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add sorryhyun/anima_lora --skill soup -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/soup, .gemini/skills/soup, .github/skills/soup and .opencode/skills/soup in your project.

What does Soup need to run?

Going by SKILL.md and its folder, Soup needs the command-line tools its instructions call (make).

Does Soup access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Soup safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Soup use?

Soup is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Soup use?

About 728 tokens (SKILL.md is roughly 2.9k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Soup?

Skills that share tags, products or a category with Soup: Init (asgeirtj/system_prompts_leaks, 69k stars), Init (asgeirtj/system_prompts_leaks, 69k stars), Init (brycewang-stanford/Auto-Empirical-Research-Skills, 4.6k stars) and Agent Coordinator Swarm Init (ruvnet/ruflo, 74k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Soup?

sorryhyun (a GitHub user) maintains it in sorryhyun/anima_lora, which has 125 GitHub stars. The repository holds 12 skills in this directory. The repository was last updated on October 10, 2026.

Source: sorryhyun/anima_lora on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.