Phewsh Skills Atlas
Tested ways to think, build, verify, and hand off—plus a trusted map of the AI-skills frontier. Phewsh curates provenance, compatibility, evidence, and continuity instead of competing on catalog size.
Phewsh Originals
Each card names its activation, supported harnesses, provenance, and evidence. “Tested” means a dated check exists—not a blanket security guarantee.
Turn an unclear request into durable Project, Next, Work, and Record context that survives a change of AI tool.
Phewsh Original
phewsh ambient on
Move one repository from unfinished to verified and shippable, one bounded slice and one honest handoff at a time.
Phewsh Original
mkdir -p ~/.claude/skills/boat-to-shore && curl -fsSL https://raw.githubusercontent.com/cleverIdeaz/phewsh-cli/main/skills/boat-to-shore/SKILL.md -o ~/.claude/skills/boat-to-shore/SKILL.md
Choose which repository receives the next execution window without losing portfolio-level state.
Phewsh Original
mkdir -p ~/.claude/skills/portfolio-boat-loop && curl -fsSL https://raw.githubusercontent.com/cleverIdeaz/phewsh-cli/main/skills/portfolio-boat-loop/SKILL.md -o ~/.claude/skills/portfolio-boat-loop/SKILL.md
Add concise think-first, simplicity, surgical-change, and goal-driven coding guidance to project agent files.
Adapted from Andrej Karpathy's published coding guidance; upstream license review remains required.
phewsh pack install karpathy-style
Plan high, execute wide, and judge high by routing work according to intelligence, taste, cost, and effort.
Phewsh Synthesis of established model-routing practices.
phewsh pack install model-routing
Audit whether an AI-assisted project remains goal-oriented, evidence-backed, governable, and transferable.
Inspired by Daniel Miessler's public AI audit prompts; no upstream text is reproduced.
phewsh pack install governance-audits
Frontier Map
Phewsh links to creators and upstream repositories. A skill, a directory, a scanner, and a complete operating system are not the same thing.
Composable skills
Focused workflows you can combine with Phewsh without adopting a whole operating system.
Composable agent loops with explicit checks and stopping conditions.
Composable engineering skills for requirements, TDD, diagnosis, research, architecture, review, and handoffs.
Complete workflow systems
Broader methodologies and harnesses. Useful to compare, but not interchangeable with a single skill.
A complete spec-driven planning, execution, research, debugging, and verification workflow.
An opinionated collection of CEO, design, engineering, QA, review, browser, and shipping workflows.
An agent harness with subagents, memory, sandboxes, and skills.
A separate agent harness mapped as a compatibility candidate.
A coherent software-development methodology built from composable planning, TDD, subagent, review, and verification skills.
A full lifecycle method for ideation, product planning, architecture, implementation, and specialist agents.
Security and auditing
Tools and collections that can touch sensitive surfaces. Every entry begins unreviewed.
A large security-focused community skill collection that requires deliberate scope review.
A security scanner for AI-agent skills; a scan is evidence, not a guarantee.
Specialist and domain work
Media, OCR, voice, and code-intelligence systems for deliberately scoped projects.
A persistent code-intelligence MCP server for repository memory and lookup.
A local or GPU OCR candidate for image and PDF workflows.
An open-source macOS video editor with an MCP surface.
An agentic video-production system with pipelines, tools, and skills.
HTML-to-video rendering for agent-produced demos and media.
An open-source AI voice studio for audio workflows.
Standards, official collections, and discovery
Primary references and broad directories. Phewsh maps them instead of copying their catalogs.
The open format specification for portable Agent Skills.
Official examples and production references for Agent Skills.
OpenAI's official catalog of skills for Codex.
Official web-development, design, deployment, and optimization skills from Vercel.
A broad discovery index for the open-source skills ecosystem; Phewsh links and filters rather than mirroring it.
A community-curated index of official and community Agent Skills.
On the workbench
These are the next compact first-party skills. They become installable only after a real SKILL.md and verification evidence exist.
Reduce cognitive load in UX, product copy, instructions, and documentation.
Phewsh Original — planned, not yet installable.Reconcile local intent, cloud identity, repository state, and generated harness context.
Phewsh Original — planned, not yet installable.Create truthful local, verify, and ship scripts from commands the repository already establishes.
Phewsh Original — planned, not yet installable.Use the strongest reasoning for planning and judgment while delegating bounded execution.
Phewsh Synthesis — planned, not yet installable.Turn an ambitious objective into evidence-gated builder and independent-critic phases with durable closure.
Inspired by Matt Shumer's Gauntlet Loop. Phewsh adds persistent project truth, authority boundaries, evidence gates, completion criteria, and Project · Next · Work · Record closure. It is not Matt Shumer's official skill.Packs
Skill describes the intellectual artifact. Pack describes how Phewsh previews, installs, links, and removes it.
The Atlas is the canonical editorial surface for originals, upstream work, provenance, compatibility, and evidence.
Open the CLI installation viewphewsh pack reads the same catalog. Managed blocks are previewed and removable; linked systems remain upstream.
phewsh pack
Build a Skill
A core skill should usually fit on one screen. Progressive disclosure keeps situational examples, scripts, templates, and rubrics in referenced files.
Use the open Agent Skills format: a directory with SKILL.md, YAML name and description, plus optional scripts, references, and assets.