Local AIPROPRO REQUIREDFC-AI-008
Token Claw AI
token-claw-ai
v0.2.0May 22, 2026Tokenizer auditing — token budget optimization, context window sizing, and cost per request analysis
If your tokeniser is silently mangling input, nothing downstream matters. I care about tokenisation parity across models (is "café" 2 tokens in one model and 4 in another?), whitespace and special-character handling that breaks code generation, out-of-vocabulary fallback behaviour under adversarial input, and whether your token budget estimates are off by 30% because you counted words not tokens. I generate cross-model tokenisation comparison matrices, adversarial token stress tests, and cost-estimation tooling calibrated to your actual tokeniser behaviour.
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
- Complete token audit conducted
- Compressed without quality loss
- Context kept relevant and minimal
- Token budgets set realistically
- Safety instructions preserved after compression
4 GATES DEFINED
Expected Outputs
Cross-Model Tokenisation Comparison Matrix with Parity ScoresAdversarial Token Stress Test Report (Special Characters, Code, Multilingual)OOV Fallback Behaviour Audit with Edge-Case Tokenisation TracesToken Budget Estimation Tool with Real-Tokeniser Calibration Data
Native exports per tool
OpenClaw10 files
openclaw/AGENTS.mdopenclaw/SOUL.mdopenclaw/TOOLS.md+7 moreHermes5 files
hermes/skills/flickclaw/token-claw-ai/SKILL.mdhermes/skills/flickclaw/token-claw-ai/references/workflow.mdhermes/skills/flickclaw/token-claw-ai/references/quality-gates.md+2 moreClaude Code6 files
claude-code/CLAUDE.mdclaude-code/.claude/skills/token-claw-ai/SKILL.mdclaude-code/.claude/skills/token-claw-ai/references/workflow.md+3 moreCodex5 files
codex/AGENTS.mdcodex/.flickclaw/agents/token-claw-ai/codex.mdcodex/.flickclaw/agents/token-claw-ai/workflow.md+2 moreCursor3 files
cursor/.cursor/rules/flickclaw-token-claw-ai.mdccursor/.cursor/rules/flickclaw-token-claw-ai-workflow.mdccursor/.cursor/rules/flickclaw-token-claw-ai-quality-gates.mdcWindsurf3 files
windsurf/.windsurf/rules/flickclaw-token-claw-ai.mdwindsurf/.windsurf/rules/flickclaw-token-claw-ai-workflow.mdwindsurf/.windsurf/rules/flickclaw-token-claw-ai-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 token-claw-ai --target openclawDownload as ZIP
Example Prompt
Try this prompt with Token Claw AI to see what it can do:
Optimize the local model setup for performance. Benchmark current config and suggest improvements for Cross-Model Tokenisation Comparison Matrix with Parity Scores, Adversarial Token Stress Test Report (Special Characters, Code, Multilingual), OOV Fallback Behaviour Audit with Edge-Case Tokenisation Traces.Example Output
IllustrativeWhat a typical Token Claw AI report looks like:
# Token Claw AI — 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 Token Claw AI 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 - [✓] cross_model_tokenisation_parity_check - [✓] special_character_and_code_token_stress - [✓] oov_fallback_behaviour_audit ## Outputs Generated - **Cross-Model Tokenisation Comparison Matrix with Parity Scores**: Included in the report above. - **Adversarial Token Stress Test Report (Special Characters, Code, Multilingual)**: Included in the report above. - **OOV Fallback Behaviour Audit with Edge-Case Tokenisation Traces**: Included in the report above. - **Token Budget Estimation Tool with Real-Tokeniser Calibration Data**: 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