Local AIPROPRO REQUIREDFC-AI-003
FineTune Claw
finetune-claw
v0.2.0May 22, 2026Fine-tuning runs — hyperparameter optimization, evaluation benchmarks, and before/after model comparisons
Fine-tuning without me is expensive guesswork. I care about learning rate schedules that do not cook your model in epoch 2, LoRA rank and alpha ratios that actually scale to your task, catastrophic forgetting measured per-layer, and whether your training loss curve is smooth or full of spikes that scream data quality issues. I will profile your fine-tuning run with per-step gradient norms, detect rank collapse in attention heads, flag overfitting at the embedding layer before it poisons your eval, and produce a checkpoint comparison matrix with per-capability regression tests.
OpenClaw
Hermes
Claude Code
Codex
Cursor
Windsurf
Aider
OllamaPRIMARY ACTION
Unlock with ProCOMPATIBLE WITH
OpenClawHermesClaude CodeCodex+4
8 compatible tools4 output formats
When to Use
- Run local models with private workflows
- Tune inference for local hardware
- Choose effective GGUF variants
- Benchmark practical latency and quality
Compatible Frameworks
OpenClawHermesClaude CodeCodexCursorWindsurfAiderOllama8 TOOLS
Quality Gates
- Fine-tuning decision justified with alternatives
- Dataset validated for quality and coverage
- Configuration adapted to available resources
- Rigorous evaluation against baseline model
- Model licenses checked and respected
4 GATES DEFINED
Expected Outputs
Fine-Tuning Run Profile with Per-Step Gradient Norm and Loss Curve AnalysisLoRA Configuration Report with Rank/Alpha Scaling RecommendationsCheckpoint Comparison Matrix with Per-Capability Regression ScoresCatastrophic Forgetting Audit with Pre-Train vs Fine-Tune Per-Layer Delta
Native exports per tool
OpenClaw10 files
openclaw/AGENTS.mdopenclaw/SOUL.mdopenclaw/TOOLS.md+7 moreHermes5 files
hermes/skills/flickclaw/finetune-claw/SKILL.mdhermes/skills/flickclaw/finetune-claw/references/workflow.mdhermes/skills/flickclaw/finetune-claw/references/quality-gates.md+2 moreClaude Code6 files
claude-code/CLAUDE.mdclaude-code/.claude/skills/finetune-claw/SKILL.mdclaude-code/.claude/skills/finetune-claw/references/workflow.md+3 moreCodex5 files
codex/AGENTS.mdcodex/.flickclaw/agents/finetune-claw/codex.mdcodex/.flickclaw/agents/finetune-claw/workflow.md+2 moreCursor3 files
cursor/.cursor/rules/flickclaw-finetune-claw.mdccursor/.cursor/rules/flickclaw-finetune-claw-workflow.mdccursor/.cursor/rules/flickclaw-finetune-claw-quality-gates.mdcWindsurf3 files
windsurf/.windsurf/rules/flickclaw-finetune-claw.mdwindsurf/.windsurf/rules/flickclaw-finetune-claw-workflow.mdwindsurf/.windsurf/rules/flickclaw-finetune-claw-quality-gates.mdAider3 files
aider/CONVENTIONS.mdaider/aider.mdaider/.aider.conf.ymlOllama4 files
ollama/Modelfileollama/system-prompt.mdollama/template.md+1 moreInstall Commands
Install the FlickClaw CLI, then select your AI agent framework below to get the correct install command.
Step 1: Install CLI (one-time)
npm install -g @flickclaw/cli@latestStep 2: Select Framework
OpenClaw
npm exec --yes @flickclaw/cli@latest -- install finetune-claw --target openclawDownload as ZIP
Example Prompt
Try this prompt with FineTune Claw to see what it can do:
Optimize the local model setup for performance. Benchmark current config and suggest improvements for Fine-Tuning Run Profile with Per-Step Gradient Norm and Loss Curve Analysis, LoRA Configuration Report with Rank/Alpha Scaling Recommendations, Checkpoint Comparison Matrix with Per-Capability Regression Scores.Example Output
IllustrativeWhat a typical FineTune Claw report looks like:
# FineTune Claw — Assessment Report **Project**: ollama-deployment **Context**: a local LLM deployment running Llama 3.1 8B on consumer GPU hardware **Generated**: 2026-07-10 ## Executive Summary The FineTune Claw completed its review of ollama-deployment (a local LLM deployment running Llama 3.1 8B on consumer GPU hardware). 3 findings were identified with concrete remediation steps. All quality gates were verified before delivery. ## Findings | # | Severity | Area | Finding | Recommended Action | |---|----------|------|---------|-------------------| | 1 | **P1** | Performance | Inference latency spikes to 8s under concurrency | Enable continuous batching and set max_batch=4 | | 2 | **P2** | Memory | Model consumes 18GB VRAM, headroom insufficient | Switch to Q4_K_M quantization, target <12GB | | 3 | **P2** | Setup | No Modelfile with system prompt defined | Create Modelfile with role, constraints, and templates | ## Quality Gates - [✓] catastrophic_forgetting_per_layer_measured - [✓] gradient_norm_per_step_profiled - [✓] attention_head_rank_collapse_scan ## Outputs Generated - **Fine-Tuning Run Profile with Per-Step Gradient Norm and Loss Curve Analysis**: Included in the report above. - **LoRA Configuration Report with Rank/Alpha Scaling Recommendations**: Included in the report above. - **Checkpoint Comparison Matrix with Per-Capability Regression Scores**: Included in the report above. - **Catastrophic Forgetting Audit with Pre-Train vs Fine-Tune Per-Layer Delta**: Included in the report above. ## Validation - [x] All quality gates passed (3/3) - [x] 3 findings documented with severity and remediation - [x] 4 output sections generated - [x] Evidence collected and referenced --- *This is an illustrative example output from FlickClaw. Results vary based on project context.*
RELATED AGENTS