AgentSkillsCN

learned-patterns

管理已学习的模式——列出、查看、归档、提升或降低其可信度。适用于想要了解 Claude 学习了哪些模式、评估模式的有效性，或对模式库进行管理时。

SKILL.md
--- frontmatter
name: learned-patterns
description: Manage learned patterns - list, view, archive, boost or penalize confidence. Use when you want to see what patterns Claude has learned, review pattern effectiveness, or manage the pattern library.
allowed-tools: Read, Write, Edit, Bash, Glob, Grep
infer: false

Learned Patterns Management

Manage the auto-learning pattern library. View, list, archive, and adjust confidence.

Quick Commands

code
/learned-patterns                    # List all active patterns
/learned-patterns list [category]    # List patterns (optionally filtered)
/learned-patterns list --low|--high  # Filter by confidence
/learned-patterns view <id>          # View pattern details
/learned-patterns archive <id>       # Archive a pattern
/learned-patterns boost <id>         # Increase confidence (+10%)
/learned-patterns penalize <id>      # Decrease confidence (-15%)
/learned-patterns stats              # Show pattern statistics

Script Execution

ActionScript
Listnode .claude/skills/learned-patterns/scripts/list-patterns.cjs [category] [--low|--high]
Viewnode .claude/skills/learned-patterns/scripts/view-pattern.cjs <id>
Archivenode .claude/skills/learned-patterns/scripts/archive-pattern.cjs <id> [reason]
Boost/Penalizenode .claude/skills/learned-patterns/scripts/adjust-confidence.cjs <id> boost|penalize
Statsnode .claude/skills/learned-patterns/scripts/pattern-stats.cjs

Pattern Lifecycle

code
DETECTION (UserPromptSubmit)
  └── Confidence: 40% (implicit) or 80% (explicit /learn)

CONFIRMATION
  └── User confirms → saved | User rejects → discarded

INJECTION (SessionStart/PreToolUse)
  └── Relevant patterns injected (max 5, ~400 tokens)

FEEDBACK LOOP
  └── Followed → +5% | Ignored → -10%

DECAY & PRUNING
  └── 30 days unused → decay | Below 20% → auto-archived

Confidence Thresholds

RangeBehavior
80-100%Always injected when relevant
50-79%Injected with context match
30-49%Injected only on strong match
20-29%Candidate for review
< 20%Auto-archived

Pattern Schema

yaml
id: pat_abc123
category: backend
type: anti-pattern | best-practice | preference | convention
trigger:
  keywords: [validation, exception]
  file_patterns: ["*CommandHandler.cs"]
content:
  wrong: "throw new ValidationException()"
  right: "return PlatformValidationResult.Invalid()"
  rationale: "Framework uses result pattern, not exceptions"
metadata:
  source: explicit-teaching | user-correction
  confidence: 0.75
  first_seen: 2025-01-10
  occurrences: 3
  confirmations: 2
  conflicts: 0

Troubleshooting

  • •Not injecting: Check confidence, file pattern match, archived status
  • •Too many injecting: Archive irrelevant patterns, review --low confidence list
  • •Conflicts with docs: Pattern auto-blocked; update docs or archive pattern

Related

  • •/learn - Teach a new pattern
  • •/code-patterns - View static code patterns documentation

IMPORTANT Task Planning Notes

  • •Always plan and break many small todo tasks
  • •Always add a final review todo task to review the works done at the end to find any fix or enhancement needed