# Skill Router

> Find and load on-demand skills from a Prawn Skills Git clone on any supported OS. Use for unfamiliar tooling, specialized tasks, skill discovery, or requests involving documents, media, cloud platforms, writing, or skill maintenance.

- Skill: `hongyime/skill-router-2` (Agent Skill)
- Install (CLI): `npx get-skill add hongyime/skill-router-2`
- Raw SKILL.md: https://api.skillmd.com/api/skills/hongyime/skill-router-2/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- License: MIT
- Author: hongyime (https://skillmd.com/u/hongyime)
- Updated: 2026-11-16
- Page: https://skillmd.com/skills/hongyime/skill-router-2

---


# Skill Router for Standalone Installs

Read `~/.config/theprawnskills/library.json` and use its `library` path as the
repository root. If absent and this skill is a symlink, resolve it to locate
the clone. For a copied installation, inspect the intended clone and repair
the pointer through the installer; the installed copy is not the library.
The repo root contains `default-profile.toml` and `skills/`.

Read `<root>/platforms/linux/README.md` for platform conventions. Search names
and descriptions with `rg` in `<root>/skills` and `<root>/platforms/linux/skills`.
Choose the smallest relevant set, prefer a standalone variant when present, and
read each selected `SKILL.md` fully before acting. Resolve symlinks before
following relative references. Never install the full library just to search it.

| Request | Starting points |
|---|---|
| API compatibility and serialized provider/consumer contracts | api-contracts; openapi-to-typescript for generated types |
| Understand an unfamiliar repository | codebase-onboarding; codebase-design for choosing changes |
| Record or find architectural decisions and rationale | architecture-decision-records; cross-harness-state for current task state |
| Bounded repair loops, agent roles, terminal/worktree coordination | bounded-agent-loop, cli-agent-router |
| Create, add, update, audit, remove skills | Standalone variants of skill-create, skill-add, skill-update, skill-cleanup, skill-remove, skill-authoring |
| Coding, review, debugging | build, check, refactor, requesting-code-review, systematic-debugging |
| Completion evidence across build/types/lint/tests | verification-loop; check remains SPEC drift only |
| Python testing; React component testing; browser flows | python-testing; react-testing; webapp-testing |
| Pi-hole, resolver health and Tailscale DNS routing | homelab-pihole-dns, docker-expert |
| Visual explanations, diagrams, UML/C4, flows, comparisons, HTML reports | visual-explainer; postplan-upload for hosted delivery |
| Documents, spreadsheets, presentations | pdf, docx, xlsx, pptx |
| Frontend design | frontend-design, web-design-guidelines, shadcn |
| Cloudflare | cloudflare, wrangler, workers-best-practices |
| Azure / Supabase / Vercel | azure-prepare, supabase, deploy-to-vercel |
| Images, video, audio | ai-image-generation, image-to-video, text-to-speech |
| Writing and marketing | technical-blog-writing, case-study-writing, content-repurposing |
| Session continuity across agents | cross-harness-state; session-handoff; molt |
| Repo standardization and compliance | repo-standardization; shell |

Verify candidate availability and tool prerequisites from the actual files.
Keep source skills and installed links/copies separate. Use the Python installer
for local exposure, and do not run OneDrive propagation or `dotagents sync`.

Keep `visual-explainer` in the daily profile with `postplan-upload`. Use its
shared dark/cyan template whenever a visual helps, including short answers.
The document order is visual overview, details, next steps. Hosting is a
separate delivery step governed by the user's authorization.

