Notre avis
Cette compétence permet d'enregistrer de nouvelles connaissances acquises pendant une session en les ajoutant aux fichiers de compétences pertinents.
Points forts
- Améliore la base de connaissances au fil du temps
- Capture le contexte et les solutions spécifiques aux problèmes
- Structure les apprentissages en catégories claires
- Automatise la révision des apprentissages avec /eod
Limites
- Nécessite une discipline pour capturer régulièrement
- Peut entraîner une duplication si les connaissances existent déjà ailleurs
- Dépend de la structure de dossier des compétences existante
Utilisez cette compétence après avoir résolu un problème ou découvert un nouveau motif de configuration que vous souhaitez retenir.
Ne l'utilisez pas pour des correctifs ponctuels sans réutilisation potentielle ou pour des informations déjà bien documentées.
Analyse de sécurité
SûrThe skill provides instructions for capturing learned knowledge into markdown files, using git for version control. It does not involve any external command execution beyond file editing and git, and poses no risk of exfiltration, destruction, or safety bypass.
Aucun point d'attention détecté
Exemples
/add-skill - I learned how to fix the PostgreSQL connection timeout by adjusting max_connections in the database skill.We just discovered that setting broadcast storm control to 100 kpps prevents HA Pi network drops. Add this to the infrastructure skill under UniFi configuration./eod - review the session for new learnings and add them to the appropriate skill files.Add Skill Knowledge
Capture newly learned information and add it to the appropriate skill knowledge base.
Purpose
When you learn something new during a session (troubleshooting techniques, configuration patterns, tool usage), this command captures that knowledge and adds it to the relevant skill for future reference.
Usage
User invokes: /add-skill
Then specify:
- What was learned - Brief description
- Skill to update - Which skill this belongs to (infrastructure, scpi-automation, database, etc.)
- Category - What section to add it to (patterns, troubleshooting, configuration, tools, etc.)
Instructions
Step 1: Identify What Was Learned
Ask yourself:
- What problem did we just solve?
- What configuration/pattern did we discover?
- What tool behavior did we learn?
- What best practice emerged?
Step 2: Determine Target Skill
Available Skills:
infrastructure- Proxmox, Docker, networking (UniFi), storagescpi-automation- Test equipment control and automationdatabase- PostgreSQL, homelab_db, ccpm_db operations[future skills as needed]
Step 3: Add to Skill Knowledge
Read the skill's SKILL.md file and add the new knowledge to the appropriate section:
Common Sections:
- Patterns - Reusable configuration patterns, templates
- Troubleshooting - Solutions to specific problems
- Configuration - Specific settings and their meanings
- Tools - Tool usage patterns and best practices
- Checklists - Pre/post task verification steps
Step 4: Format the Addition
Use this format:
### [Topic Name] (LEARNED: YYYY-MM-DD)
**Context:** [What problem/situation led to this learning]
**Solution/Pattern:**
```
[Configuration, command, or pattern]
```
**Why This Works:**
- [Explanation point 1]
- [Explanation point 2]
**When to Use:**
- [Use case 1]
- [Use case 2]
**Related:**
- [Links to docs, GitHub issues, session summaries]
Example Usage
User: /add-skill
You respond:
I'll add the knowledge we just gained. What did we learn?
Recent session context:
- HA Pi network failure troubleshooting
- UniFi storm control configuration
- STP/Loop Protection settings
Which should I capture?
User: "The UniFi port configuration for storm control and loop protection"
You:
Adding to: .claude/skills/infrastructure/SKILL.md
Section: UniFi Port Configuration
Category: Troubleshooting > Network Stability
[Shows the new section being added]
Auto-Capture During /eod
The /eod command now includes a step to review the session and identify learnings:
## Skill Knowledge Updates
During this session, we learned:
- [ ] UniFi storm control configuration (pKts/s thresholds)
- [ ] Loop Protection = STP Edge Port in UniFi
- [ ] Medical-critical network hardening patterns
Should I add these to skills? (Y/n)
Skill Knowledge Principles
What to Capture:
✅ Capture:
- Configuration patterns that solved real problems
- Tool behaviors discovered through troubleshooting
- Best practices learned through experience
- Settings and their actual effects (not documentation copy)
- Troubleshooting procedures that worked
❌ Don't Capture:
- One-off fixes specific to single issue
- Information already well-documented elsewhere
- Temporary workarounds
- Obvious/common knowledge
Quality Over Quantity:
-
Be specific: "Storm control: Broadcast 100 kpps prevents HA Pi network drops"
-
Not generic: "Storm control prevents network problems"
-
Include context: "Loop Protection enabled on uplink ports prevents STP blocking during topology changes"
-
Not vague: "Enable loop protection"
Skill File Locations
.claude/skills/
├── infrastructure/
│ └── SKILL.md # Proxmox, Docker, UniFi, networking
├── scpi-automation/
│ └── SKILL.md # Test equipment patterns
├── database/
│ └── SKILL.md # PostgreSQL, migrations, queries
└── [new-skill]/
└── SKILL.md # Created as needed
Creating New Skills
If knowledge doesn't fit existing skills:
-
Create new skill directory:
.claude/skills/[skill-name]/ -
Create SKILL.md with structure:
# [Skill Name] Skill **Skill Name:** [skill-name] **Version:** 1.0.0 ## Purpose [What this skill covers] ## Allowed Tools [Tools this skill can use] ## Patterns [Learned patterns and configurations] ## Resources [Documentation references] -
Document in CLAUDE.md - Add to skills list
Verification
After adding skill knowledge:
- [ ] Knowledge added to appropriate SKILL.md section
- [ ] Includes context (what problem it solved)
- [ ] Includes specific values/commands (not just concepts)
- [ ] Dated with (LEARNED: YYYY-MM-DD)
- [ ] Committed with descriptive message
- [ ] Cross-referenced to session summary or GitHub issue
Git Commit Pattern
git add .claude/skills/[skill-name]/SKILL.md
git commit -m "skill: Add [topic] to [skill-name] knowledge base
Learned during [context/issue].
Added:
- [Specific pattern/config 1]
- [Specific pattern/config 2]
Solves: [Problem description]
Reference: [Session summary/issue link]
"
Example: Adding today's UniFi learning:
git add .claude/skills/infrastructure/SKILL.md
git commit -m "skill: Add UniFi storm control and STP configuration patterns
Learned during HA Pi network failure investigation (Issue #36).
Added:
- Storm control thresholds (pKts/s format)
- Loop Protection = STP Edge Port mapping
- Medical-critical network hardening patterns
Solves: Network drops during HA Pi service restarts
Reference: docs/session-summary-2026-01-10.md, Issue #36
"
HomeLab Agent - Continuous Learning System Version: 1.0
Generateur de Documentation API
Documentation
Genere automatiquement de la documentation API OpenAPI/Swagger.
Rédacteur Technique
Documentation
Rédige de la documentation technique claire selon les meilleurs style guides.
Système de formulaires de documentation typés
Documentation
Utilisez la syntaxe `(doc ...)` pour ajouter des annotations typées, des descriptions, des tâches (todo) et d'autres métadonnées directement dans le code Scheme. Les annotations sont extractibles via des commandes comme lf-todo et lf-types, et s'intègrent au vérificateur de types, où les déclarations de type dans les doc prennent le pas sur l'inférence. Idéal pour documenter les fonctions, marquer des déprécations ou lister des améliorations localisées sans recourir à un système externe.