similar-resources
Given a Japanese NLP GitHub repo or Hugging Face model/dataset (URL / owner/repo / tool name), find repositories or models/datasets that do the same or related processing. Mines the bundled dataset for content-similar items, then expands via web research across both GitHub and Hugging Face, then merges and re-ranks.
npx skills add taishi-i/awesome-japanese-nlp-resources --skill similar-resources --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Given a seed resource (GitHub URL, Hugging Face URL, owner/repo, or tool/model name), it searches a bundled dataset for related items, expands the search across Web resources on GitHub and Hugging Face, and merges and re-ranks results.
How it works
- Establish current year values and print seeds in the targeted workflow (Step 0 and Step 1 patterns).
- Validate input; require a seed resource and provide usage examples if missing.
- Normalize the seed to identify seed forms (GitHub URL, Hugging Face URL, owner/repo, or tool/model name).
- Locate data files using RESOURCES_PATH from the plugin data and a data file named resources.json.
- Run an inline Python script that loads the resources.json, builds a fingerprint for the seed, and scores all other items by matching category, semantic labels, and description tokens with IDF-weighted scoring.
- If the seed is not found in the local dataset, optionally WebFetch or WebSearch the seed to characterize it and derive 4–6 English stem keywords representing the task and tool. Re-run a local similarity pass in keyword mode to recover closest local items.
- Derive web-search angles using the seed’s task and category, generating 6–8 queries (with a bias toward Japanese content) across GitHub and Hugging Face, including at least 2 Hugging Face queries.
- Use WebSearch + WebFetch to extract and collect URLs for GitHub repos, Hugging Face models, and Hugging Face datasets, filtering out seed and avoiding deeper paths. Deduplicate within each kind and report only unlisted, web-found resources not already in the dataset.
When to use it
Use when the user names a SPECIFIC repository, model, or tool and wants alternatives, equivalents, or related resources. Trigger phrases include various Japanese phrases for similarity and alternatives, or English phrases like "alternatives to fugashi". For topic searches, use the separate find-new-resources capability.
What it can touch
- Uses allowed-tools: Bash, WebSearch, WebFetch.
- Interacts with the local resources.json data file to compute similarity.
- Performs web-based discovery and categorizes results as github, hf_model, or hf_dataset.
Caveats
- The approach relies on the presence and structure of resources.json and the declared workflow steps.
- Only resources not already in the dataset are reported from the web search portion.
- File and tool names are referenced exactly as in the skill description (e.g.,
resources.json,WebFetch,WebSearch).
Find repositories similar to: "$ARGUMENTS" — first from the bundled awesome-japanese-nlp-resources dataset, then from the web. ## Instructions ### Preamble — Establish the current date Before anything else, run this once and remember the values — every step that mentions a year refers to them: ```bash echo "YEAR_NOW=$(date +%Y)" echo "YEAR_PREV=$(($(date +%Y) - 1))" ``` Substitute these wherever this skill writes `${YEAR_NOW}` or `${YEAR_PREV}`. **Do not hardcode years.** ### Step 0 — Validate input This skill needs a seed repository, model, or tool. If `$ARGUMENTS` is empty or blank, **stop immediately** and output: ``` Usage: /awesome-japanese-nlp-resources:similar-resources <github-url | huggingface-url | owner/repo | tool-name> Examples: /awesome-japanese-nlp-resources:similar-resources mecab /awesome-japanese-nlp-resources:similar-resources fugashi /awesome-japanese-nlp-resources:similar-resources https://github.com/polm/cutlet /awesome-japanese-nlp-resources:similar-resources WorksApplications/Sudachi /awesome-japanese-nlp-resources:similar-resources manga-ocr /awesome-japanese-nlp-resources:similar-resources https://huggingface.co/cl-nagoya/ruri-large Pass the repository/mod
- Instructions
- Preamble — Establish the current date
- Step 0 — Validate input
- Step 1 — Normalize the seed
- Step 2 — Locate the data files
- Step 3 — Locate the seed and score similarity (inline Python)
- Step 4 — If the seed is not in the dataset
- Step 5 — Derive web-search angles
- Step 6 — Web research
- Step 7 — Re-rank
- Step 8 — Format the output
- Step 9 — Edge cases
echo "YEAR_NOW=$(date +%Y)" echo "YEAR_PREV=$(($(date +%Y) - 1))" echo "RESOURCES_PATH=$RESOURCES_PATH" github candidate grep -iqE "github\.com/<owner>/<repo>[\"/]" "$RESOURCES_PATH" && echo "in list — drop" || echo "unlisted — keep" hf_model candidate grep -iqE "huggingface\.co/<owner>/<name>[\"/]" "$RESOURCES_PATH" && echo "in list — drop" || echo "unlisted — keep" hf_dataset candidate grep -iqE "huggingface\.co/datasets/<owner>/<name>[\"/]" "$RESOURCES_PATH" && echo "in list — drop" || echo "unlisted — keep"
What does the similar-resources skill do?
Given a Japanese NLP GitHub repo or Hugging Face model/dataset (URL / owner/repo / tool name), find repositories or models/datasets that do the same or related processing. Mines the bundled dataset for content-similar items, then expands via web research across both GitHub and Hugging Face, then merges and re-ranks.
How do I install it?
Run `npx skills add taishi-i/awesome-japanese-nlp-resources --skill similar-resources --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From taishi-i/awesome-japanese-nlp-resources, a repository with 1,000 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.