AgentSkillsCN

knowledge

展示知识库状态与近期学习成果

SKILL.md
--- frontmatter
name: knowledge
description: Display knowledge base status and recent learnings
user_invokable: true

Knowledge

Display the current state of the project's knowledge base and recent learnings.

What This Does

Shows:

  • •Learning mode status (on/off)
  • •Knowledge base statistics (entry counts per category)
  • •Recent learnings extracted
  • •Cache statistics

Instructions

  1. •Read knowledge/state.json for learning mode status
  2. •Read each knowledge file and count entries:
    • •knowledge/cache/classifications.md
    • •knowledge/learnings/patterns.md
    • •knowledge/learnings/quirks.md
    • •knowledge/learnings/decisions.md
  3. •Extract recent entries (last 5) from learnings files
  4. •Format and display

Output Format

code
╔═══════════════════════════════════════════════════╗
║           Project Knowledge Base                   ║
╚═══════════════════════════════════════════════════╝

📚 Learning Status
───────────────────────────────────────────────────
Mode: ON (since 2026-01-08 14:00)
Last Extraction: 5 minutes ago
Extractions This Session: 3

📊 Knowledge Statistics
───────────────────────────────────────────────────
Cache:
  - Classification entries: 23

Learnings:
  - Patterns: 8 entries
  - Quirks: 3 entries
  - Decisions: 5 entries
  - Total: 16 insights

📝 Recent Learnings
───────────────────────────────────────────────────
[Pattern] "Use async/await for API calls in this codebase"
  Discovered: 2026-01-08 | Confidence: high

[Quirk] "Auth module uses non-standard token format"
  Discovered: 2026-01-07 | Confidence: high

[Decision] "Chose Redis over in-memory cache for session storage"
  Made: 2026-01-06 | Confidence: high

💡 Commands
───────────────────────────────────────────────────
/learn      - Extract insights now
/learn-on   - Enable continuous learning
/learn-off  - Disable continuous learning

When Knowledge Base is Empty

code
╔═══════════════════════════════════════════════════╗
║           Project Knowledge Base                   ║
╚═══════════════════════════════════════════════════╝

📚 Learning Status
───────────────────────────────────────────────────
Mode: OFF
No extractions yet

📊 Knowledge Statistics
───────────────────────────────────────────────────
Knowledge base is empty.

💡 Get Started
───────────────────────────────────────────────────
Use /learn to extract insights from your current session.
Use /learn-on to enable continuous learning.

The knowledge base will grow as you work, capturing:
  - Patterns that work well in this project
  - Quirks and gotchas to remember
  - Decisions and their rationale

Steps

  1. •Read knowledge/state.json
  2. •Read frontmatter from each knowledge file to get entry counts
  3. •Parse recent entries from learnings files (look for ## Pattern:, ## Quirk:, ## Decision: headers)
  4. •Format and display the summary
  5. •If files are missing or empty, show the "empty" state

Notes

  • •Entry counts come from frontmatter entry_count field or by counting ## headers
  • •Recent learnings are shown most recent first (by discovered/made date)
  • •This is a read-only command - it doesn't modify any files