Skills may execute instructions and code that could affect your environment. Marketplace scans reduce risk but do not guarantee safety. Always review files, run your own security checks, and use at your own risk.
self-improving-agent
Security Scan Summary
Status: Safe
Source: Syntic Skills registry
Automated security scan completed with no high-risk patterns detected. Manual review is still required.
About This Skill
Use when curating auto-memory into durable project knowledge. Analyze patterns, promote learnings to rules, extract solutions into skills, review memory health.
Downloadable SKILL.md
Download SKILL.md and place it in your Syntic skills folder. For Syntic Code, install in your local skills directory, review contents, and run in a controlled environment first. Acknowledge the risk notice above to enable the download.
--- name: self-improving-agent description: Use when curating auto-memory into durable project knowledge. Analyze patterns, promote learnings to rules, extract solutions into skills, review memory health. category: Engineering version: 1.0.0 tools: [] --- # Self-Improving Agent Curate Claude Code's auto-memory into durable project knowledge and enforced rules. ## Commands | Command | Purpose | |---------|----------| | `/si:review` | Analyze MEMORY.md—find promotion candidates, stale entries, consolidation opportunities | | `/si:promote` | Graduate pattern from MEMORY.md → CLAUDE.md or `.claude/rules/` | | `/si:extract` | Turn proven pattern into standalone skill | | `/si:status` | Memory health dashboard—line counts, topics, recommendations | | `/si:remember` | Explicitly save important knowledge to auto-memory | ## Memory Architecture **Where things live:** | File | Writes | Scope | Load Behavior | |------|--------|-------|----------------| | `./CLAUDE.md` | You + /si:promote | Project rules | Full file | | `~/.claude/CLAUDE.md` | You | Global preferences | Full file | | `~/.claude/projects/<path>/memory/MEMORY.md` | Claude (auto) | Project learnings | First 200 lines | | `~/.claude/projects/<path>/memory/*.md` | Claude (overflow) | Topic-specific notes | On demand | | `.claude/rules/*.md` | You + /si:promote | Scoped rules | When matching files | **Promotion lifecycle:** 1. Claude discovers pattern → auto-memory (MEMORY.md) 2. Pattern recurs 2-3 times → /si:review flags as candidate 3. User approves → /si:promote graduates to CLAUDE.md or rules/ 4. Pattern becomes enforced rule, not just a note 5. MEMORY.md entry removed → frees space for new learnings ## Core Concepts **Auto-memory is capture, not curation.** Excellent at recording discoveries, but has no judgment about: - Which learnings are temporary vs. permanent - Which patterns should become enforced rules - When 200-line limit wastes space on stale entries - Which solutions are good enough to become reusable skills That's what this plugin does. **Promotion = graduation.** Moves from scratchpad (MEMORY.md) to rules (CLAUDE.md or `.claude/rules/`). - MEMORY.md: "I noticed this project uses pnpm" (context) - CLAUDE.md: "Use pnpm, not npm" (enforced) **Rules directory for scoped knowledge.** Not everything belongs in CLAUDE.md. Use `.claude/rules/` for patterns applying only to specific file types: ```yaml paths: - "src/api/**/*.test.ts" - "tests/api/**/*" --- - Use supertest for API endpoint testing - Mock external services with msw - Always test error responses, not just happy paths ``` ## Agents **memory-analyst**: Analyzes MEMORY.md to identify: - Entries recurring across sessions (promotion candidates) - Stale entries referencing deleted files - Related entries needing consolidation - Gaps between MEMORY.md learnings and CLAUDE.md enforcement **skill-extractor**: Generates complete skill from proven pattern: - SKILL.md with proper structure - Reference documentation - Examples and edge cases - Ready for plugin install or clawhub publish ## Hooks **error-capture (PostToolUse → Bash)**: Monitors command output for errors. When detected, appends structured entry to auto-memory with: - Failed command - Error output (truncated) - Timestamp and context - Suggested category Token overhead: Zero on success, ~30 tokens only when error detected.
Bundle Download
Includes SKILL.md and bundled support files where provided. Risk acknowledgement is required.
Install Targets
Syntic App
- 1. Create a dedicated folder for this skill in your local skills library.
- 2. Place SKILL.md into that folder.
- 3. Restart Syntic and invoke this skill on matching tasks.
Syntic Code (CLI)
- 1. Save SKILL.md in your local Syntic Code skills directory.
- 2. Keep related files in the same skill folder.
- 3. Run in a safe environment and validate outputs.
Source
https://github.com/alirezarezvani/claude-skills/blob/main/engineering-team/self-improving-agent/skills/self-improving-agent/SKILL.md
Open Source LinkRelated Skills
a11y-audit
Use when auditing WCAG 2.2 Level A/AA accessibility, fixing violations in React, Next.js, Vue, Angular...
Engineeringadversarial-reviewer
Use when reviewing recent code changes or a PR before merge and you want a genuinely critical review, not...
Engineeringagent-designer
Use when architecting multi-agent systems, selecting orchestration patterns, or evaluating agent performance.
Engineeringagent-harness
Use when building bounded agentic loops with verified task execution and state machines.