Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

Back to skills

Evolving Loop

BSecurity

Self-Evolving Development Loop - Dynamic skill generation with learning and evolution

81 stars
0 votes
0 copies
0 views
Added 2/7/2026
ai-agentsbash

Security Analysis

B80/100
highPerforms destructive filesystem operations
criticalSends environment variables or credentials to an external URL

Scanned 2/12/2026

Install to Claude Code

$npx -y skills add claude-world/director-mode-lite --skill evolving-loop --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Evolving Loop?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Evolving Loop
[![Security: B — Skills Directory](https://www.skillsdirectory.com/api/skills/claude-world-evolving-loop/badge)](https://www.skillsdirectory.com/skills/claude-world-evolving-loop)

More formats (shields.io, HTML) on the badges page.

Download Zip
Files
SKILL.md
---
name: evolving-loop
description: Self-Evolving Development Loop - Dynamic skill generation with learning and evolution
user-invocable: true
---

# Self-Evolving Development Loop

Execute an autonomous development cycle that dynamically generates, validates, and evolves its own execution strategy. Integrates with Meta-Engineering memory system for pattern learning and tool evolution.

> **Architecture Details**: See [docs/EVOLVING-LOOP-ARCHITECTURE.md](../../../docs/EVOLVING-LOOP-ARCHITECTURE.md)

---

## Usage

```bash
# Start new task
/evolving-loop "Your task description

Acceptance Criteria:
- [ ] Criterion 1
- [ ] Criterion 2
"

# Flags
/evolving-loop --resume    # Resume interrupted session
/evolving-loop --status    # Check status
/evolving-loop --force     # Clear and restart
/evolving-loop --evolve    # Trigger manual evolution
/evolving-loop --memory    # Show memory system status
```

---

## How It Works

```
┌──────────────────────────────────────────────────────┐
│  8-Phase Self-Evolving Loop                          │
├──────────────────────────────────────────────────────┤
│                                                      │
│  Phase -2: CONTEXT_CHECK  → Check token pressure     │
│  Phase -1A: PATTERN_LOOKUP → Match task patterns     │
│                                                      │
│  ┌─────────────── Main Loop ───────────────┐        │
│  │ Phase 1: ANALYZE   → Extract AC         │        │
│  │ Phase 2: GENERATE  → Create skills      │        │
│  │ Phase 3: EXECUTE   → TDD implementation │        │
│  │ Phase 4: VALIDATE  → Score 0-100        │        │
│  │ Phase 5: DECIDE    → SHIP/FIX/EVOLVE    │        │
│  │ Phase 6: LEARN     → Extract patterns   │        │
│  │ Phase 7: EVOLVE    → Improve skills     │        │
│  └──────────────────────────────────────────┘        │
│                                                      │
│  Phase -1C: EVOLUTION → Update memory (on SHIP)     │
│                                                      │
└──────────────────────────────────────────────────────┘
```

---

## Execution

When user runs `/evolving-loop "$ARGUMENTS"`:

### 1. Handle Flags

```bash
STATE_DIR=".self-evolving-loop"
MEMORY_DIR=".claude/memory/meta-engineering"
CHECKPOINT="$STATE_DIR/state/checkpoint.json"

# --status: Show current state
if [[ "$ARGUMENTS" == *"--status"* ]]; then
    /evolving-status
    exit 0
fi

# --memory: Show memory system status
if [[ "$ARGUMENTS" == *"--memory"* ]]; then
    echo "Memory System Status:"
    if [ -d "$MEMORY_DIR" ]; then
        echo "Tool Usage: $(jq '.tools | length' "$MEMORY_DIR/tool-usage.json" 2>/dev/null || echo "0") tools"
        echo "Patterns: $(jq '.task_patterns | keys | length' "$MEMORY_DIR/patterns.json" 2>/dev/null || echo "0") patterns"
        echo "Evolution: v$(jq -r '.version' "$MEMORY_DIR/evolution.json" 2>/dev/null || echo "0")"
    else
        echo "(Not initialized - will create on first run)"
    fi
    exit 0
fi

# --resume: Continue from checkpoint
if [[ "$ARGUMENTS" == *"--resume"* ]]; then
    if [ ! -f "$CHECKPOINT" ] || [ "$(jq -r '.status' "$CHECKPOINT")" == "idle" ]; then
        echo "No active session to resume."
        exit 1
    fi
fi

# --force: Clear old state
if [[ "$ARGUMENTS" == *"--force"* ]]; then
    rm -rf "$STATE_DIR/state/*" "$STATE_DIR/reports/*" "$STATE_DIR/generated-skills/*"
fi
```

### 2. Initialize (First-Run Safe)

```bash
# Create directories (first-run safe)
mkdir -p "$MEMORY_DIR"
mkdir -p "$STATE_DIR"/{state,reports,generated-skills,history,backups}

# Helper: Read JSON with fallback
read_json_safe() {
    local file="$1"
    local default="$2"
    if [ -f "$file" ]; then
        cat "$file" 2>/dev/null || echo "$default"
    else
        echo "$default"
    fi
}

# Detect first run
IS_FIRST_RUN=false
if [ ! -f "$MEMORY_DIR/patterns.json" ]; then
    IS_FIRST_RUN=true
    echo "📝 First run detected - initializing memory system..."
fi

# Initialize memory files if missing (see docs for full schema)
```

### 3. Delegate to Orchestrator

**CRITICAL**: Use context isolation - orchestrator runs in fork context.

```markdown
Task(subagent_type="evolving-orchestrator", prompt="""
Request: $ARGUMENTS
Task Type: $TASK_TYPE (from pattern matching)

Execute phases in sequence, each in fork context.
Return only brief status updates (1 line per phase).
Store ALL detailed output in files.

Return format:
📊 CONTEXT: [OK/Warning] - [N]% usage
🔍 PATTERNS: Matched [type], [N] recommendations
✅ ANALYZE: [N] AC identified
✅ GENERATE: Created v[N] skills
🔄 EXECUTE: Iter [N] - [status]
✅ VALIDATE: Score [N]/100
➡️ DECIDE: [SHIP/FIX/EVOLVE]
""")
```

---

## Output Example

```
🚀 Starting Self-Evolving Loop (Meta-Engineering v2.0)...

📊 CONTEXT: OK - 15% usage
🔍 PATTERNS: Matched 'auth', 3 recommendations
✅ ANALYZE: 5 acceptance criteria identified
✅ GENERATE: Created executor-v1, validator-v1, fixer-v1
🔄 EXECUTE: Iteration 1 - 4 files modified, 3/5 tests passing
✅ VALIDATE: Score 72/100
➡️ DECIDE: FIX (minor test failures)
🔄 EXECUTE: Iteration 2 - 2 files modified, 5/5 tests passing
✅ VALIDATE: Score 94/100
➡️ DECIDE: SHIP
📚 LEARN: 2 patterns identified
🧬 EVOLUTION: Updated memory
✅ SHIP: All criteria met!

📊 Summary: 2 iterations, 6 files changed, 5/5 AC complete
```

---

## Phase Agents

| Phase | Agent | Output File |
|-------|-------|-------------|
| ANALYZE | `requirement-analyzer` | `reports/analysis.json` |
| GENERATE | `skill-synthesizer` | `generated-skills/*.md` |
| EXECUTE | (generated executor) | codebase changes |
| VALIDATE | (generated validator) | `reports/validation.json` |
| DECIDE | `completion-judge` | `reports/decision.json` |
| LEARN | `experience-extractor` | `reports/learning.json` |
| EVOLVE | `skill-evolver` | evolved skills |

---

## State Files

```
.self-evolving-loop/          ← Session state (temporary)
├── state/checkpoint.json     ← Current state
├── reports/*.json            ← Phase outputs
├── generated-skills/*.md     ← Dynamic skills
└── history/*.jsonl           ← Event logs

.claude/memory/meta-engineering/  ← Persistent memory
├── tool-usage.json           ← Usage statistics
├── patterns.json             ← Learned patterns
└── evolution.json            ← Evolution history
```

---

## Stop / Resume

```bash
# Stop after current phase
touch .self-evolving-loop/state/stop

# Resume later
/evolving-loop --resume
```

---

## Related

- [/evolving-status](../evolving-status/SKILL.md) - View status and memory
- [evolving-orchestrator](../../agents/evolving-orchestrator.md) - Phase coordinator
- [Architecture Details](../../../docs/EVOLVING-LOOP-ARCHITECTURE.md) - Full technical docs

Attribution

claude-worldclaude-world
View sourceMore from claude-world →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".

1066601 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

686011 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

651 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →