Lallapallooza

@lallapallooza source repo

14 skills
0 followers
11 hours ago last updated

▌
Vs Core Rfc · lallapallooza bundle
RFC and feature design pipeline from concept to implementation spec. Use this skill when the user wants to design a new feature end-to-end, write an RFC, create a technical specification, or go from idea to implementation plan. Also use when the user says "RFC", "design a feature", "spec this out", "technical design", "feature design", or describes a substantial new capability that needs design before implementation.

0
installs
▌
Vs Core Arch · lallapallooza bundle
Architecture analysis and interface design. Use this skill when the user needs to analyze existing architecture, design a new system or module, compare architectural approaches, or evaluate tradeoffs. Also use when the user says "architecture", "design", "how should we structure", "interface design", "module design", "design it twice", "what's the right abstraction", or discusses system-level concerns.

0
installs
▌
Vs Core Init · lallapallooza bundle
Create or extend a project-root CLAUDE.md. Two modes. INIT writes a new seed file after a short interview plus a silent probe; auto-selected when no CLAUDE.md exists. APPEND inserts a single learning (a gotcha, a command, a common flow, a knowledge-drift note) into the existing CLAUDE.md at the right section; auto-selected when CLAUDE.md exists and the user has a specific thing to add. Use when the user says "init this project", "generate a CLAUDE.md", "/vs-core-init", "add this to CLAUDE.md", or when a session surfaces a learning worth persisting.

0
installs
▌
Vs Core Audit · lallapallooza bundle
Adversarial review with parallel specialized agents. Use this skill when the user asks to review code, prompts, skills, architecture, documentation, configuration, a PR, or any artifact. Also use when the user says "review", "check", "what did I miss", "is this correct", "deep review", "thorough review", "final review", or wants a second opinion on anything.

0
installs
▌
Vs Core Debug · lallapallooza bundle
Systematic debugging with root cause analysis. Use this skill when the user has a bug, test failure, crash, or unexpected behavior and needs to find the root cause. Also use when the user says "debug", "why is this failing", "it's broken", "this doesn't work", "fix this bug", or when a test fails unexpectedly.

0
installs
▌
Vs Core Grill · lallapallooza bundle
Socratic interview to reach shared understanding before starting work. Use this skill when the user wants to discuss a plan, design, or approach before implementing. Also use when the user says "grill me", "interview me", "let's think through this", "what should we consider", or when starting a complex task where requirements are unclear.

0
installs
▌
Vs Core Plain · lallapallooza bundle
Write or rewrite text in plain, simple English using ASD-STE100 and plain-language rules. Use this skill when the user asks to simplify text, says "simple english", "STE", "ASD-STE100", "plain language", "make it simpler", "too verbose", "oververbose", "I don't get the point", or is drafting a message for a non-native-English audience such as upstream maintainers, mailing lists, issue threads, or docs.

0
installs
▌
Vs Core Tropes · lallapallooza bundle
Review text for AI writing tropes. Use this skill when the user wants to check prose for AI-sounding patterns, or says "tropes", "check for AI slop", "does this sound AI", "deslop", or wants to make text sound more human.

0
installs
▌
Vs Core Autoloop · lallapallooza bundle
Generate a single generic pasteable autonomous iteration loop prompt/runbook. Use when the user wants a long-running optimization, coverage, evaluation, research, cleanup, triage, game-tuning, parity, performance, latency, fuzzing, cost, documentation, dependency, alert, or security loop that can keep working without repeated human babysitting while each iteration has an explicit verdict.

0
installs
▌
Vs Core Research · lallapallooza bundle
Deep multi-source technical research. Use this skill when the user asks to research a technology, compare alternatives, investigate how something works in the codebase, make a tech decision, understand a topic deeply, or needs a thorough technical briefing. Also use when the user says "look into", "investigate", "compare", "what should we use for", or "how does X work".

0
installs
▌
Vs Core Implement · lallapallooza bundle
Full implementation pipeline with planning, spec-driven verification, review gates, and bidirectional spec communication. Use this skill when the user wants to implement a feature, build something substantial, or execute a plan with quality gates. Also use when the user says "implement", "build this", "develop", "execute the plan", or describes a feature that requires multiple files and careful engineering.

0
installs
▌
Vs Core Interactive · lallapallooza bundle
Conversational companion for day-to-day coding work. Loads standing behavioral guidelines (careful thinking, minimal code, surgical changes, goal-driven verification) plus anti-sycophancy interaction style, a verification iron law, and banned hedge-phrases. Automatically loads matching shared references (rationalization-rejection, self-critique protocol, trust-boundary) and language/perf judgment files when it detects the relevant domain. Suggests structured vs-core-* sub-skills (grill, research, rfc, implement, audit, debug, tropes) when the user's task matches their scope. Optionally writes session logs to .spec/ for non-trivial work. Use when the user wants a responsive working partner rather than a pipeline phase.

0
installs
▌
Vs Core Profile Amd · lallapallooza bundle
CPU profiling on AMD Zen processors, uProf-centric with perf/samply/likwid/bpftrace as first-class complements. Use when the user wants to profile, benchmark, or microarch-analyze native code (C/C++/Rust/Go) on AMD Ryzen/EPYC hardware. Also use when the user says "uprof", "AMDuProf", "IBS", "Instruction-Based Sampling", "Zen PMU", "amd profile", "AMD timechart", "AMDuProfPcm", "roofline on AMD", "profile on Ryzen", "profile on EPYC", "zen 4 tma", "amd top-down", or asks about AMD-specific counter events, cache/branch/TLB/memory-bandwidth/power measurement on Zen.

0
installs
▌
Vs Core Profile Nvidia · lallapallooza bundle
GPU profiling on NVIDIA hardware, Nsight-centric (Nsight Systems + Nsight Compute) with PyTorch profiler / Kineto / NVTX / CUPTI / DCGM / Meta HTA / Perfetto as first-class complements. Use when the user wants to profile, benchmark, or microarch-analyze CUDA kernels, ML training, or LLM inference on NVIDIA GPUs (Pascal through Blackwell). Also use when the user says "nsys", "ncu", "Nsight Systems", "Nsight Compute", "PC Sampling", "CUPTI", "Hopper TMA", "wgmma", "warp specialization", "tensor core utilization", "achieved DRAM bandwidth", "SM occupancy", "stall reasons", "NCCL profile", "NVLink topology", "MFU", "torch.profiler", "Kineto", "NVTX", "DCGM", "B200 profile", "FP8 profiling", "FP4 profiling", "roofline on H100", "wave quantization", "GPU idle time", "data loader bottleneck".

0
installs

14 published skills