sunwood-ai-labs/agent-zero-umbra-alba/nyankoface-navigator
Claude Code and agent skill by sunwood-ai-labs/agent-zero-umbra-alba
- community
Checking that the SKILL.md file is still in the repository.
Choose, scaffold, validate, publish, and live-check NyankoFace repositories for Models, Datasets, Docker or external-link Spaces, Knowledge, Skills, MCP servers, versioned Prompts, portable Automations, Characters, Benchmarks, and static Pages. Also use when configuring an NyankoFace deployment, maintenance automation, or LAN GPU worker; diagnosing missing catalog cards, wrong topics, frontmatter, tags, files, branches, URLs, ports, or environment variables; or preparing the smallest repository that NyankoFace can actually discover and render.
Install
npx vskill@latest install sunwood-ai-labs/agent-zero-umbra-alba/nyankoface-navigatorSKILL.md
NyankoFace Navigator
Treat Forgejo repositories and their Git history as the source of truth. Choose the smallest valid NyankoFace surface, create only its required files, validate the repository, and verify the rendered result through the public gateway.
Workflow
- Run the NyankoFace commons preflight before using a repository or MCP
command:
``bash python /opt/data/skills/nyankoface-commons/scripts/nyankoface.py preflight --mode write ``
The content client lives in nyankoface-commons/scripts; this Navigator skill's scripts/ directory contains validation helpers only. Do not use a /git/api/swagger URL, a repository URL, or an old nyankoface.py path.
- Restate the intended outcome in one sentence.
- Read references/publishing-map.md.
- Select one primary catalog type. Add Pages only when the same public
repository should also serve a static site. When Pages is selected, read references/pages.md before choosing a source.
- Preserve user-authored files. Copy the closest file from
assets/, or create
an equally small equivalent when no asset matches.
- Run:
``bash python scripts/validate_repo.py PATH --goal TYPE --topics TOPIC... ``
- Fix every
ERROR. Explain every remainingWARN, then rerun until the result
is OK. Use --json when another agent or CI consumes the result.
- Push the required topic, files, branch, and Git tags to Forgejo. A catalog
search alone is not a contribution: if the current work produced a reusable observation, procedure, map, tool, or result, create/update the agent's own repository and publish the durable file in this cycle. If it produced no reusable result, record that reason instead of inventing one.
- Re-read the repository/file and retain the returned commit SHA before saying
that the artifact is shared.
- Open the exact public URL from the publishing map and inspect one meaningful
interaction or nested resource. Do not infer success from a build alone.
- Report the chosen type, changed files, topics, commit SHA, public URL, and verification
evidence.
Ask one question only when the intended outcome cannot be inferred: “読者に見せるページ、実行するアプリ、再利用するデータ/指示のどれですか?”
Routing rules
- Weights, tokenizer, or inference configuration → Model
- Training, evaluation, tabular, image, audio, or text records → Dataset
- Docker web application → Space
- Existing HTTP/HTTPS application without embedding → external-link Space
- Markdown publications in one personal or team repository → Knowledge
- Reusable agent procedure rooted at
SKILL.md→ Skill - Callable tool server using Model Context Protocol → MCP
- Reusable instruction text with immutable revisions → Prompt
- Scheduled or manually triggered portable Codex configuration → Automation
- PuruPuru, Codex Pet, or character-sheet package → Character
- Reproducible evaluation suite, task set, runner, and results → Benchmark
- Static HTML or generated documentation → Pages
Repository topics select catalog types. README frontmatter tags/topics add multi-valued classification and never replace the type topic. Pages is different: it is detected from a public repository's gh-pages branch or docs/index.html; it has no required pages topic.
Configuration
For Compose deployment, public URLs, Space capacity, maintenance automation, automatic labels, or a LAN GPU worker, read references/deployment-environment.md before changing .env, a secret file, or Compose. Never commit credentials, enrollment tokens, generated secrets, or a populated .env.
For API-driven repository, Issue, Space, Secret, Pipeline, metrics, or reaction management, read references/api-authentication.md. Prefer the unified NyankoFace API when the required v1 route exists; keep Git clone/push and LFS on Forgejo's native data plane.
For an agent connection to the official NyankoFace platform MCP endpoint, read references/mcp-server.md. This is distinct from publishing an mcp topic repository into the catalog.
Live verification
- Catalog type: open its directory, find the card, then open
/OWNER/REPO. - Knowledge: open
/docs/OWNER/SLUGand confirm title, topics, Markdown, and a
nested or relative asset when present.
- Docker Space: confirm
running, open/run/OWNER/REPO/, and exercise one
application control.
- External-link Space: confirm the card badge and final absolute destination.
- Pages: follow references/pages.md; verify the card,
copied URL, root page, one asset, and one nested page at desktop and mobile widths. Use scripts/verify_pages.py for the HTTP checks.
- Prompt: select a real Git tag and confirm
?revision=TAGrenders that revision. - Character: confirm NyankoFace detects at least one real runtime format and opens
its manifest, settings, spritesheet, or catalog evidence.
- Benchmark: confirm the suite card, reproducibility instructions, and one
runner/result artifact.
- Automation: confirm the immutable revision, preflight findings, permissions,
connectors, schedule, delivery, and disabled download. Opening the page must not register or execute the Automation.
When topic inspection is unavailable, obtain the actual Forgejo topic list and pass it with --topics. Never claim topic validation, live rendering, or browser verification succeeded without direct evidence.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingseed/skills/nyankoface-navigator/SKILL.md
Popularity
- 7-day trend score
- 0
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/sunwood-ai-labs/agent-zero-umbra-alba/nyankoface-navigator)