Installs into .claude/skills of the current project.
Sponsors
SSkills DirectorySkills Directory
Know which skills are safe — weekly.
SSkills DirectorySkills Directory
Is that skill safe to install?
Are you the author of Coding Skill Selector?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/axgord-coding-skill-selector)
---
name: coding-skill-selector
description: Select and load coding skills by file extensions and domains
---
## Coding skill selector
Load coding skills for the languages/domains THIS TASK will touch.
1. `Skill("preferences")` + `Skill("architecture")` — EXCEPT on a read-only
route. The gate prompt's stack line names the active workflow
(`[N] explore @ load_skills`, `[N] debugging @ load_skills`): those two
never write code, and both skills only govern WRITING (style, abstraction
thresholds), so on them skip this step, and skip any personal
`preferences-<lang>` skill from rule 3 too (keep `lang-<x>` and tool
skills — they are needed to READ code correctly). `investigate`,
`planning`, review and every coding route load the full set.
2. **Language menu** (preferred path): the session's routing prompt included a
"Project language menu" block — an engine-run scan with per-extension file
counts and tool probes. Pick from it ONLY the languages this task will
actually touch: counts are a hint (1 stray file ≠ project language), the
task text decides. Do NOT re-scan the project yourself.
No menu in this session (spawned agent, scanner unavailable, empty block) →
fallback: determine extensions from the task's files yourself.
Tool-bound skills (a skill that mandates a CLI tool) load ONLY when the
language's menu line ends in an available-probe marker — `(<tool> OK)`
vs `(<tool> missing)` — or, when there is no menu or the language's line
is absent from it, after one `command -v <tool>` probe succeeds. Tool
unavailable → skip the tool skill and work with regular tools; a tool
discipline only applies where the tool exists.
3. By file extension (menu-picked or fallback):
- .py → `Skill("lang-python")`
- .hx → `Skill("lang-haxe")`, plus `Skill("hxq")` per the tool-bound rule
above (`hx … (hxq OK)`, or `command -v hxq` succeeds)
- .as → `Skill("lang-as3")`
<!-- Add your own language mappings, e.g.: -->
<!-- - .cs → `Skill("preferences-csharp")` -->
<!-- - .cpp, .h → `Skill("preferences-cpp")` -->
<!-- - .c → `Skill("preferences-c")` -->
<!-- Pair language skill with personal preferences skill if you have one (e.g. `preferences-haxe`) -->
4. **Top-up rule**: if mid-task you touch a file of a language whose skills
are not loaded (task turned out wider than it looked) — load that
language's skills per rule 3, including its tool conditions, BEFORE
editing the file.
5. By domain (detect from imports, paths, or task description):
- Pony (`-lib pony`, `pony.xml`, `pony.*` imports, `@:listen`/`@:auto`/`@:own` metas) → `Skill("domain-pony")`
- YOLO / object detection → `Skill("domain-yolo")`
- Pixi.js (imports from `pixi.js`, PIXI globals) → `Skill("domain-pixi")`
- Person re-identification (ReID) → `Skill("domain-reid")`
- Game dev (physics, precision, camera, sprites) → `Skill("domain-gamedev")`
- UI components / display-tree apps (ownership, state, events) → `Skill("domain-ui")`
- Diagnosing a bug / fixes that don't stick after 2+ rounds → `Skill("debugging")`
- OpenFL / hxcpp native target → `Skill("target-openfl-native")`
6. By tooling/platform:
- CMakeLists.txt / CMake build → `Skill("build-cmake")`
- .github/workflows / CI / GitHub Actions → `Skill("ci-github-actions")`
- AWS Lambda / .NET deploy → `Skill("aws-lambda")`
- MCP server setup / .mcp.json / MCP troubleshooting → `Skill("mcp-setup")`
- Claude Code settings.json / hooks / permissions / plugin config → `Skill("claude-code-config")`
- Workflow YAML in templates/ or .claude/workflows/ → `Skill("wf-yaml-authoring")`
- Setting up / tuning a project linter, or a new project with no lint
config → `Skill("lint-setup")` (skip if not installed; it loads the
user's `preferences-lint` lists itself — neither loads for normal
coding/review)
- Fetching web docs / external content → `Skill("web-reading")`
- Verifying a web/canvas app via Playwright or a browser MCP (screenshots, captures, perf) → `Skill("browser-verify")`
7. If 3+ independent files or parallel work → `Skill("task-delegation")` —
EXCEPT when you are a spawned sub-agent (the gate prompt's stack shows
`subagent` at the root): a worker never spawns agents of its own, so the
skill is dead weight there — skip it.
Call `Skill(name)` for EVERY matched skill.
**Gap check**: If files involve a language/domain/tool with no matching skill above, ask the user: "No skill for [X] — proceed without, or create one first?"