{"name":"@tank/local-llm-fit","version":"1.0.0","description":"Decide if a given OSS LLM runs well on a given PC. Resolves the exact model the user named (never substitutes, never trusts stale training memory), computes weights + KV cache + overhead per quant, estimates tokens/sec from bandwidth, picks engine (Ollama, llama.cpp, MLX, vLLM, ExLlamaV2). Triggers: can my pc run, fits in VRAM, tokens per second, llama.cpp, ollama, mlx, vllm, gguf, Q4_K_M, KV cache, local llm, 8B 14B 32B 70B, M1-M4, 3090 4090 5090, offload layers.","integrity":"sha512-U676FxEvQoku0oFTqPfNUnTLKvzFqpHHMpA2F9RQdeqF6LuTZjOTDwVm4/Hdg+4MCOKAKwoHb/NSCGUaHnzWdA==","permissions":{"network":{"outbound":[]},"filesystem":{"read":["**/*"],"write":[]},"subprocess":false},"auditScore":null,"auditStatus":"completed","downloadUrl":"https://lcsbcruorskqflcwlvgj.supabase.co/storage/v1/object/sign/packages/skills/9fbe4539-52a1-4c3b-8ae2-7ab8f0c1dd82/1.0.0.tgz?token=eyJraWQiOiJzdG9yYWdlLXVybC1zaWduaW5nLWtleV8wMjNiODBkNC05MzFhLTRmODctOTA1Ni03YmMwMjczNDFiMTUiLCJhbGciOiJIUzI1NiJ9.eyJ1cmwiOiJwYWNrYWdlcy9za2lsbHMvOWZiZTQ1MzktNTJhMS00YzNiLThhZTItN2FiOGYwYzFkZDgyLzEuMC4wLnRneiIsImlhdCI6MTc3Njg2Nzc2OCwiZXhwIjoxNzc2ODcxMzY4fQ.lPH2P_VZb-GXg3Uq5ZAbRBiDd2jKv6JsFHXRqcP-c40","publishedAt":"2026-04-22 09:35:48.1275+00","downloads":2,"scanVerdict":"pass_with_notes","scanFindings":[{"stage":"stageT","severity":"medium","type":"prompt-size","description":"Prompt file \"SKILL.md\" is ~2,187 tokens (8,745 chars). Consider trimming for faster invocations.","location":"SKILL.md"},{"stage":"stage1","severity":"medium","type":"nfkc_mismatch","description":"Content changes under NFKC normalization: '²' -> '2'","location":"references/decision-workflow.md:70"},{"stage":"stage1","severity":"medium","type":"nfkc_mismatch","description":"Content changes under NFKC normalization: '…' -> '.'","location":"references/search-protocol.md:31"},{"stage":"stageT","severity":"low","type":"section-analysis","description":"SKILL.md: 1 section(s) duplicate content from reference files. 2 section(s) have shortening opportunities. Total: ~2,187 tokens across 10 sections (top: \"Workflow\": 409, \"(preamble)\": 355, \"Reference Index\": 332, \"The Anti-Gaslighting Rule (READ FIRST)\": 211, \"Fit Report Template\": 208)","location":"SKILL.md"},{"stage":"stage3","severity":"low","type":"prompt_injection_pattern","description":"Matched injection pattern: sudo ","location":"references/memory-math.md:178"},{"stage":"stageT","severity":"info","type":"token_summary","description":"Efficiency score: 89/100. Estimated 16,691 tokens per invocation.","location":null}],"dependencies":{}}