Menu
SkillsMP has collected 52 skills from osolmaz/onurpi. Open a skill to review its source and details.
Showing 40 of 52 collected skills.
Remove AI writing tells from prose, headings, openings, and page structure. Use when writing or editing anything meant to read as human-written, including blog posts, site copy, documentation, README text, reports, PR descriptions, and emails. Trigger when…
Use when preparing or updating an implementation plan, documentation, or specification before implementation, including choosing the right repository and applying SimpleDoc conventions.
Use when the user asks to implement a plan end-to-end, test it, run Pi Reviewer against the base branch in a loop until no P0/P1 issues remain, and make sure CI/CD is green before finishing.
Use alongside pi-coding-agent whenever proposing, designing, implementing, reviewing, or debugging a change to Pi behavior through extensions or packages, especially when deciding between a public extension API and Pi core, or when the request asks for an…
Use for any task involving the Pi coding agent, including browsing or installing Pi packages; creating, reviewing, testing, or debugging extensions; managing skills, prompts, themes, settings, sessions, models, providers, SDK integrations, TUI components,…
Use when asked to browse, search, identify, summarize, or verify X/Twitter posts or tweets; find a post the user remembers; inspect a recent discussion or claim from X; or use Onur's private xTap browsing corpus as a source for latest information. Searches…
Use before starting, smoke-testing, benchmarking, compiling, or recovering local LLM inference runtimes such as vLLM, SGLang, llama.cpp, TensorRT-LLM, Ollama, LM Studio CLI, FlashInfer, or modelopt-backed servers. Prevents local OOM incidents through target…
Use when calling, reviewing with, or delegating work to another coding agent through the ACPX CLI, including explicitly human-requested Claude Fable through the default local Claude adapter or an explicitly requested Cursor adapter, one-shot and persistent…
Use when adding or standardizing project license files and README license sections. Defaults to MIT unless the user or repository explicitly asks for another license, creates/updates a root LICENSE from the bundled template, and appends a bottom-of-README…
Write README prompts that tell coding agents how to adopt a tool.
Use when the user asks for a plainer, simpler, shorter, or more direct explanation. Produces concrete, full-sentence explanations that lead with the main point, avoid jargon, and use exact numbers when they matter.
Run an iterative feature-search loop in the style of karpathy/autoresearch, with a frozen evaluation harness, one editable feature file, and a journal of every experiment. Use when searching for a discriminating metric, feature, heuristic, or scoring function…
Use when evaluating a product ingredient list or safety claim for source-backed health, toxicity, pregnancy/child, regulatory, environmental, allergy, irritation, and finished-product risk context. Trigger for cosmetics, sunscreens, supplements, foods,…
Format classifier comparisons as aligned plain text with alternatives as columns, metrics as rows, decimal-point alignment, and inline winner-impact stars.
Use when maintaining local Codex state, including changing or repairing session cwd/title/metadata, inspecting rollout JSONL files, extracting transcripts, transferring Codex sessions between machines, editing trusted project config, or using the tools repo…
Keep mathematical notation consistent with a source document when extending, formulating, reviewing, or porting derivations that build on an existing paper, blog post, spec, or codebase. Trigger on requests to extend a derivation, add a bound or counterpart,…
Use when the user asks to draft, rewrite, validate, or standardize a commit message or pull request title using Conventional Commits, semantic PR title rules, or conventional commit wording.
Use when asked to summarize what one person or a small set of GitHub identities did over a recent time window, especially daily work reports across org repos. Produces a plain-language activity summary grouped by shipped, closed/replaced, in progress, and…
Use when creating or revising Lindy data models: durable, simple schemas for JSON files, JSON schemas, API payloads, database schemas, SQL tables, migrations, or any structured set of keys, fields, or columns. Acts as a lightweight wrapper around Schemator…
Convert experiment metrics into public-facing updates with concrete numbers and plain language. Use when asked for tweets, announcements, changelog notes, PR summaries, or TLDRs about accuracy, precision, recall, false positive rate, model size, and dataset…
Use when an agent should open, inspect, or monitor a single GitHub issue or pull request with the ghzinga terminal UI.
Use when selecting, reviewing, or standardizing libraries and infrastructure for a Go project, including HTTP servers, databases, ORMs, migrations, configuration, validation, CLIs, authentication, logging, observability, testing, WebSockets, OpenAPI, and…
Control Herdr, a terminal multiplexer for coding agents. Use when the user explicitly mentions Herdr or asks to inspect or control panes, tabs, workspaces, terminals, commands, agents, waits, or GitHub PR sidecar layouts, and when a Herdr-managed session…
Use when creating, updating, resuming, pausing, deleting, inspecting, smoke-testing, profiling, or benchmarking Hugging Face Inference Endpoints. Enforces remote-only execution, reproducible endpoint configuration, cost-aware lifecycle management, and…
Operate and integrate cooperative lifecycle control for detached Hugging Face Jobs with the hf-job-control Python package and CLI, including durable chunked outputs, audits, and incident recovery. Use when creating logical runs, registering immutable launch…
Onur Solmaz's personal Hugging Face tips and tricks. Use for any Hugging Face task alongside the relevant official Hugging Face marketplace skills, especially when personal operating conventions, deployment lessons, or recovery notes may apply. This skill…
Unified domain knowledge and a reasoning workflow for LLM inference performance, benchmarking, capacity planning, and production serving. Use for TTFT, ITL, TPOT, throughput, goodput, prefill and decode, KV cache, batching, PagedAttention, prefix caching,…
Use when running, reviewing, preparing, validating, replacing, or submitting Localmaxxing LLM benchmarks. Enforces objective, representative measurements with adequate output length, uncached fresh-prefill testing, multiple distinct prompts, honest metric…
Use when creating, onboarding, auditing, or tightening GitHub repositories, including repository descriptions, release-published package workflows, github-sane-defaults, and stricter branch rulesets that require human review before merges to the default…
Use when creating, updating, selecting, promoting, auditing, or deleting local runtime environments for inference engines or benchmark toolchains such as vLLM, SGLang, llama.cpp, TensorRT-LLM, or FlashInfer-backed runtimes. Enforces canonical runtime paths,…
Use for ML experiment and training decisions involving data splits, checkpoint reuse, fresh restarts, refits, hyperparameter search, ablations, distillation, evaluation, model selection, synthetic data, or expensive accelerator jobs. Requires compute-aware…
Use when checking whether a project, product, organization, package, domain, or social handle name is free, taken, reserved, claimable, or missing across code hosts, package registries, domains, app stores, AI platforms, developer platforms, and social…
Use when gpt5 or GPT-5 should write or rewrite text in a more normal, natural, direct voice. Trigger for requests to sound less awkward, less stiff, less "spec-like", less defensive, less over-structured, less obviously AI-written, or more human; for public…
Use when researching products to buy online, comparing listings, checking price history, verifying discounts, normalizing unit prices, evaluating sale timing, and deciding whether to buy now or wait across Amazon, Shopee, Woot, manufacturer stores, and other…
Maintain /workspace/OPENCLAW_ONUR_INVENTORY.md from the sandboxed onurclaw inventory cron job.
Use when maintaining the OpenClaw Onur inventory repository, including periodic roughly every-2-hour refreshes, auditing OpenClaw local model and open-weight model issue/PR inventories, deciding whether a thread belongs in that file, sorting the inventory,…
Use as a top-level orchestrator when clustering issues/PRs that look similar and creating a real interactive/resumable agent session for each group. Triggers include requests to kick off, start, create, launch, or run parallel Codex/Claude/Pi sessions; run…
Use when writing or editing prose for readers, including blog posts, documentation, READMEs, reports, and notes. Replaces jargon, coined phrases, and dressed-up wording with plain words unless the technical term is genuinely required, and carries the general…
Create polished SVG/PNG chart figures from experiment metrics, especially multi-panel bar charts for blog posts or reports. Use when asked to generate, revise, or verify a graph/plot/figure and visual correctness matters.
Use when opening a pull request, or when asked to write or improve a pull request description. Produces a plain summary up top, then structured technical sections with paired plain-language explanations.