Notre avis
Cette compétence permet de charger des compétences générées automatiquement dans la session Claude Code en cours sans redémarrage.
Points forts
- Active immédiatement une compétence sans redémarrer la session.
- Prend en charge la correspondance partielle du nom de la compétence.
- Liste les compétences disponibles via une commande simple.
Limites
- Les compétences chargées ne persistent pas entre les sessions.
- Nécessite que le registre des compétences soit à jour.
- Fonctionne uniquement avec Claude Code.
Utilisez cette compétence lorsque vous souhaitez utiliser une compétence générée automatiquement pendant la session en cours sans interrompre votre flux de travail.
Ne l'utilisez pas si vous avez besoin que la compétence soit disponible de manière permanente ; installez-la plutôt via Skills CLI.
Analyse de sécurité
PrudenceThe skill enables loading external skill content and runs npx commands that could execute arbitrary code from unverified community packages. While not inherently destructive, it introduces a risk of unintended code execution if used carelessly.
- •Uses npx to install community skills, which could execute untrusted code from npm packages.
- •Instructs AI to treat loaded skill content as active instructions, potentially opening a vector for malicious skill injection.
Exemples
List all auto-generated skills currently available in my session.Load the skill named 'grep-edit-workflow' into the current session.name: load-skill description: Load an auto-generated skill into the current session for immediate use arguments:
- name: skill-name description: "Name of the skill to load (partial match supported)" required: false
Load Auto-Generated Skill
This command loads a previously generated skill into the current session, making it immediately active without requiring a restart.
How It Works
Skills in Claude Code are normally discovered at session start. This command bypasses that limitation by:
- Retrieving the skill content from the registry
- Outputting it with clear delimiters that signal "these are active instructions"
- Making the skill immediately available for the current conversation
Usage
List Available Skills
If no skill name is provided, list all available auto-generated skills:
auto-skill discover
Display the output to the user so they can choose which skill to load.
Search for Community Skills
If the user wants to find community skills, use Skills CLI:
npx skills find <query> # Interactive search
npx skills add owner/repo # Install skill
Load a Specific Skill
When a skill name is provided:
auto-skill load "<skill-name>"
IMPORTANT: The output from the auto-skill load command contains the skill instructions wrapped in clear delimiters (======). You should:
- Display this output directly to the user
- Treat the content between the delimiters as active instructions you should follow when relevant
- Confirm to the user that the skill is now loaded
Example Output
======================================================================
SKILL LOADED: grep-edit-workflow-a1b2c3
Confidence: 85%
Tokens: ~500
======================================================================
## Instructions
1. Search for the relevant patterns in the codebase
2. Read the relevant file(s) to understand the current state
3. Make the necessary edits to the file
======================================================================
END OF SKILL INSTRUCTIONS - FOLLOW WHEN RELEVANT
======================================================================
Notes
- The skill registry is automatically updated when skills are created/deleted
- Use
--rebuildflag if the registry seems out of date - Skills loaded this way are session-scoped (they persist for the current conversation)
- Multiple skills can be loaded in the same session
Expert Next.js App Router
Developpement
Un skill qui transforme Claude en expert Next.js App Router.
Générateur de README
Developpement
Crée des README.md professionnels et complets pour vos projets.
Rédacteur de Documentation API
Developpement
Génère de la documentation API complète au format OpenAPI/Swagger.