kylesnowschwartz/claude-code-hero/level-7-the-skill-quest
Claude Code and agent skill by kylesnowschwartz/claude-code-hero
- community
Checking that the SKILL.md file is still in the repository.
Claude Code Hero Level 7: The Skill Quest of Doom -- create a skill directory with SKILL.md in .claude/skills/
Install
npx vskill@latest install kylesnowschwartz/claude-code-hero/level-7-the-skill-questSKILL.md
Objective
Create a skill at .claude/skills/hero-knowledge/SKILL.md that Claude activates automatically when the context matches.
Why This Matters
Commands wait to be called by name. Rules activate by file path. But skills are something different. They're knowledge that activates on its own. Claude reads the description, recognizes the moment, and summons the skill without being asked.
Think of it this way: a slash command is a spell you cast. A skill is instinct -- expertise that surfaces when the situation demands it. You don't invoke a skill. You create the conditions, and it appears.
The Quest
Beyond the cavern of tripwires, a library. Not the dusty kind. This one breathes. Scrolls drift from shelves when a question is asked, returning when the answer is given. Knowledge that serves the seeker without being summoned.
You're going to build one of these scrolls. Every scroll in this library bears a name. Yours is hero-knowledge.
Your hero has been mapping dungeons since Level 1. Time to formalize that instinct. You'll create a skill that draws ASCII dungeon maps -- a cartography scroll that activates when someone asks to map, draw, or visualize a dungeon.
A skill has three parts:
- The directory:
.claude/skills/hero-knowledge/ - The file:
SKILL.mdinside that directory - The frontmatter:
nameanddescriptionfields that tell Claude when to activate
The description field does the heavy lifting. It tells Claude when to reach for this knowledge. Write it like a trigger condition:
- Good:
"Use when the user asks to draw, map, or visualize a dungeon layout. Generates ASCII dungeon maps." - Bad:
"Dungeon stuff"
Create .claude/skills/hero-knowledge/SKILL.md with:
- YAML frontmatter with
name(matching the directory:hero-knowledge) and adescriptionthat triggers on map/dungeon/cartography requests - A body that instructs Claude how to draw an ASCII dungeon map -- rooms, corridors, a legend, the hero's position marked with
@
Here's a starting point for the body:
#####+##### #....#....# #.A..+..B.# #....#..@.# #####+#####
Try it
Two ways to test, and you should try both:
- Let Claude find it. Say "draw me a dungeon map" or "map this dungeon." Watch for Claude to invoke the Skill tool automatically. An ASCII map should appear using the conventions from your skill.
- Invoke it directly. Type
/hero-knowledgein Claude Code. The skill content loads into the conversation. Then ask for a map -- Claude will follow the instructions you wrote.
If auto-activation doesn't fire, sharpen the description. The description is the trigger -- Claude matches it against the current conversation to decide whether to load the skill.
Beyond the two required fields
name and description are the only fields you need, and they're all this quest asks for. But the frontmatter has more, and four of them change what a skill can do:
- `context: fork` -- runs the skill in a subagent instead of the current conversation. The work happens, the result comes back, and the intermediate reading never lands in your context. Pair it with
agentto choose which agent handles the fork. - `paths` -- glob patterns. The skill only loads when you're working on matching files. A skill scoped to
**/*.sqlstays out of the way until it's relevant. - `allowed-tools` -- restricts the skill to a specific set of tools while it's active. A read-only research skill has no business calling
Write. - `model` -- pins the skill to a specific model regardless of what the session is using.
There's a fifth thing that isn't a field, and it's the one that bites people: skill descriptions share a visibility budget of roughly two percent of the context window. Install enough skills and the longest descriptions get dropped from auto-discovery entirely. They still work when invoked directly with /skill-name -- they just stop surfacing on their own. A skill that used to trigger and quietly stopped usually didn't break; it got crowded out.
Hints
Hint 1
A skill is a folder inside .claude/skills/ containing a SKILL.md file:
Hint 2
The frontmatter needs name and description. The description drives auto-activation -- write it like you're telling a colleague when to open a reference doc:
Hint 3
If auto-activation doesn't fire, the description might be too vague. Try being more specific about trigger words: "draw", "map", "visualize", "dungeon layout", "cartography."
Two things sharpen a description more than length does. Write it in the third person -- "This skill should be used when the user asks to..." reads as an instruction to Claude, where "Use when..." reads as an instruction to the user. And quote the phrases a person would actually type, rather than describing the topic in the abstract.
Verification
When you're ready, run /claude-code-hero:verify to check your work.
Filesystem Check
- Path:
.claude/skills/hero-knowledge/SKILL.md
Content Check
- Frontmatter contains a
namefield - Frontmatter contains a
descriptionfield
Further Reading
- Skills -- official docs on creating skills, frontmatter configuration, and skill sharing
Connection
The scroll drifts back to its shelf. But it remembers you now. Next time someone asks for a map, it will find them.
The quest you're reading right now? It's a SKILL.md. The verification that checks your work? Also a SKILL.md. You just forged the same thing that's been guiding you through this dungeon.
Count the artifacts on your belt. A command (hero-spell). A rule (hero-protocol). A hook that connects them. And now a skill (hero-knowledge). Four components. Each one a different kind of power.
One more companion to summon before the final forge.
Security
- Security scanPassed, 95 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 repositoryCheckingskills/level-7-the-skill-quest/SKILL.md
Popularity
- Repository stars
- 86
- Repository forks
- 4
- 7-day trend score
- 22
Stars and forks are for the whole GitHub repository, not this skill alone. 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/kylesnowschwartz/claude-code-hero/level-7-the-skill-quest)