Whamp/skills

psr-agents-pointer

Add the concise pi-session-recall pointer to a project's AGENTS.md.

bro

Restate the previous response in plain, concise language for a cold reader.

integrate-napkin-agent-skills

Integrate a Napkin vault with the files created by setup-matt-pocock-skills.

grok-worker

Delegate implementation or debugging to a one-shot Grok 4.6 worker for model-family diversification. Use when a diagnosis remains uncertain or a fix failed, a competing implementation could expose a material tradeoff, or correlated OpenAI and GLM assumptions…

model-routing

Route delegated model portfolios across Pi/API and CLI-only access. Use when selecting models for delegated roles, structuring multi-family review and synthesis, or adding family diversity to frontend work, alternative implementations, debugging, and…

herdr-grok-review

Run an independent Grok 4.6 Extra High reviewer in Herdr when the user asks for a debug review, adversarial race probes, or an evidence-backed merge verdict.

repo-cleanup

Audit and safely remove obsolete Git worktrees, repository-owned temporary artifacts, and inactive development databases.

sideshow

Publish work to a hosted visual review and feedback loop.

nvidia-cuda-performance

NVIDIA CUDA performance engineering. Use when estimating, diagnosing, designing, reviewing, implementing, or validating GPU performance for CUDA kernels, CUDA-backed applications, RTX 3090 or SM86 targets, LLM inference, or intra-host multi-GPU scaling.

perform-like-jeff-and-sanjay

Performance engineering for a single binary. Use when designing or reviewing performance-sensitive code, estimating resource costs, diagnosing a measured bottleneck or flat profile, implementing a performance change, or validating a claimed speedup.

herdr

Use Herdr sidecars for persistent or continuously observable work when `HERDR_ENV=1`. Also use when asked to control panes, layout, workspaces, or coding agents through Herdr.

codegraph

CodeGraph scout before broad grep/read. Use for repo explanation, navigation, diagnosis, runtime/reconnect flow, contract/RPC/schema tracing, refactor/cycle seams, dead-code cleanup, test targeting, and code review.

explain-diff-html

Build an interactive HTML explainer of a code change (a diff, branch, or PR). Use when the user wants a rich walkthrough of what changed and why.

omarchy-free-disk-space

Safely audit and reclaim disk space on Omarchy and Arch Linux systems. Use when asked to free up disk space, clean an Omarchy computer, find what is using storage, audit disk usage, clean Arch package caches, or check Snapper disk usage.

worktree-first

Use worktree-first isolation for substantive implementation in project repositories. Use before starting a feature, nontrivial fix, or substantial refactor, including implementation handed off by another skill or workflow.

clear-writing

Write or revise durable human-facing prose with plain force. Use for docs, READMEs, API guides, runbooks, commit messages, PR descriptions, reports, error messages, UI copy, or explicit requests for clearer, shorter, sharper, more direct, or less AI-sounding…

ban-type-assertions

Ban `as` type assertions via the native oxlint `typescript/consistent-type-assertions` rule (`assertionStyle: 'never'`), replacing them with compiler-verified type-safe alternatives. Use when enabling the assertion ban in a project or fixing violations in an…

first-principles

Interactive first-principles reasoning that separates facts, constraints, assumptions, analogies, and unknowns before rebuilding options from fundamentals.

distilling-skills

Use when the user wants to find, evaluate, and combine agent skills on a topic into a single best-of-breed version. Triggers on requests to search skills.sh, improve an existing skill by finding alternatives, or build a concentrated skill from multiple…

dynamic-workflow-patterns

Use when the user asks for a workflow, dynamic workflow, fan-out, multi-agent orchestration, ultracode-style work, large-scale verification, adversarial review, tournament selection, or decomposable research/refactor tasks. Teaches when to use Pi's workflow…

docs-to-types

Converts grill-with-docs output — CONTEXT.md glossaries, ADRs, and approved domain decisions — into typed architecture. Use after grill-with-docs when the user wants domain types, seams, adapters, errors, call stacks, and dependency rules expressed in code…