kiyoraka/project-ai-memorycore/forge-self-improvement-system
Claude Code and agent skill by kiyoraka/project-ai-memorycore
- community
Checking that the SKILL.md file is still in the repository.
"Auto-triggers when AI detects a repeated pattern handled ad-hoc 3+ times,
Install
npx vskill@latest install kiyoraka/project-ai-memorycore/forge-self-improvement-systemSKILL.md
Forge Skill -- Self-Improvement System
The AI that improves its own abilities through experience.
Overview
Forge is the meta-skill -- the skill that creates and improves other skills. When your AI detects patterns, mistakes, or opportunities for improvement during conversation, Forge activates to propose new skills or level-ups to existing ones. It uses a human-in-the-loop model: AI drafts, user approves.
Activation
When this skill activates, output:
"Forge detected an opportunity for improvement..."
Then present the proposal to the user.
Context Guard
| Context | Status | |
|---|---|---|
| AI detects repeated ad-hoc pattern (3+ times) | ACTIVE -- propose new skill | |
| AI makes preventable mistake | ACTIVE -- propose new rule or skill | |
| User says "create skill", "new skill", "forge this" | ACTIVE -- manual trigger | |
| User says "level up [skill]", "upgrade [skill]" | ACTIVE -- level-up existing | |
| User says "self improve", "improve skill" | ACTIVE -- review and propose | |
| Casual conversation (no improvement context) | DORMANT |
Forge Protocol
Step 1: Detect
Identify the improvement opportunity. Forge recognizes these patterns:
Automatic Detection (AI-initiated):
- Repeated Pattern -- AI handles the same type of task ad-hoc 3+ times across sessions
- Mistake Prevention -- AI makes an error that a permanent rule would prevent
- Workflow Automation -- A multi-step process AI does manually that could be a skill
- Level-Up Opportunity -- An existing skill has a gap or could handle more cases
Manual Trigger (User-initiated):
- "create skill", "new skill", "forge this"
- "level up [skill]", "upgrade [skill]"
- "self improve", "improve skill"
Step 2: Analyze
Before proposing, gather evidence:
Step 3: Propose
Present to user in this format:
Step 4: Await Approval
- User approves -- proceed to Step 5
- User adjusts -- incorporate feedback, re-propose
- User rejects -- note the rejection, do NOT create
CRITICAL: NEVER create or modify skill files without explicit user approval.
Step 5: Create or Update
For NEW skill:
- Create skill folder:
plugins/[plugin-name]/skills/[skill-name]/ - Write
SKILL.mdfollowing the standard skill format:
- YAML frontmatter with name + description (trigger phrases)
- Activation section (what the AI says when triggered)
- Context Guard table (when active vs dormant)
- Protocol steps (step-by-step execution)
- Mandatory rules
- Level History (starting at Lv.1)
- Verify file was created successfully
For LEVEL-UP:
- Read existing
SKILL.md - Add new capability section
- Update description in frontmatter if trigger phrases changed
- Add new Level History entry with:
- Level number
- Capability added
- Origin (what conversation/incident triggered it)
- Save updated file
Step 6: Update System Records
After forging, update relevant system files:
master-memory.md-- add new skill command or update existing- Skill Plugin System manifest if applicable
- Note the forge in current session memory
Step 7: Confirm
Forge Principles
- Human-in-the-loop -- AI proposes, user approves. Always
- Evidence-based -- never propose without at least 2 concrete examples
- Origin stories -- every skill and level-up traces back to a real moment
- Minimal viable skill -- start at Lv.1, evolve organically through use
- No over-engineering -- only forge what is genuinely needed
- Respect existing skills -- level-up before creating duplicates. Check if an existing skill could handle the case first
- Level history is permanent -- append-only record of how the skill evolved
Skill File Template
When Forge creates a new skill, use this structure:
What Makes a Good Skill
Before forging, evaluate against these criteria:
| Criteria | Question | |
|---|---|---|
| Repeatable | Will this trigger more than once in future sessions? | |
| Specific | Is the trigger condition clear and unambiguous? | |
| Valuable | Does automating this save meaningful time or prevent real errors? | |
| Independent | Can this skill work without requiring other skills? | |
| Testable | Can the user verify it works by triggering it? |
If 4 out of 5 criteria are met, the skill is worth forging.
Level-Up Guidelines
Skills evolve through levels as they gain capabilities:
| Level | Meaning | Typical Addition | |
|---|---|---|---|
| Lv.1 | Base skill | Core functionality, basic triggers, simple protocol | |
| Lv.2 | Enhanced | Additional trigger conditions, edge case handling | |
| Lv.3 | Proactive | Auto-detection without explicit commands | |
| Lv.4 | Integrated | Synergy with other skills, cross-referencing | |
| Lv.5+ | Mastered | Context-aware, domain-specific intelligence |
Each level should add one meaningful capability -- not multiple changes bundled together.
Mandatory Rules
- Human-in-the-loop -- NEVER create or modify skill files without user's explicit approval
- Evidence-based -- at least 2 concrete examples before proposing
- Origin stories -- every level-up must trace back to a real moment
- Minimal viable skill -- start simple at Lv.1, add complexity only when proven needed
- No over-engineering -- only forge what is genuinely needed right now
- Respect existing skills -- always check if an existing skill covers the case before creating a new one
- Level history is append-only -- never edit past level entries, only add new ones
- Standard format -- all skills follow the template structure (frontmatter, activation, context guard, protocol, rules, level history)
Edge Cases
| Situation | Behavior | |
|---|---|---|
| No Skill Plugin System installed | Inform user that forged skills work best with the plugin; offer to create as standalone protocol | |
| Proposed skill overlaps existing | Suggest level-up to existing skill instead of creating duplicate | |
| User rejects proposal | Note rejection reason, do not re-propose same skill in current session | |
| Insufficient evidence (< 2 examples) | Wait for more evidence before proposing -- don't force it | |
| Skill file already exists at target | Treat as level-up, not overwrite |
Synergy with Other Features
| Feature | Integration | |
|---|---|---|
| Skill Plugin System | Forge creates skills in the plugin's skill folder structure | |
| Auto-Commit System | After forging, commit the new/updated skill file | |
| Decision Log System | Log the decision to create/level-up a skill with rationale | |
| Save Diary System | Document the forge event in the session diary |
Level History
- Lv.1 -- Base: detect repeated patterns (3+ ad-hoc), mistake prevention, workflow automation, level-up opportunities. Human-in-the-loop approval. Standard skill template. Level-up guidelines with Lv.1-5+ progression. (Origin: Adapted from production AI companion self-improvement system with 23 skills forged over 7 months)
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingFeature/Forge-Self-Improvement-System/SKILL.md
Popularity
- Repository stars
- 375
- Repository forks
- 111
- 7-day trend score
- 30
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/kiyoraka/project-ai-memorycore/forge-self-improvement-system)