Skills — Public

Public AI agent skills for Claude Code and compatible agent hosts.

Skills

Grouped by source directory (see Repository layout). Every skill is a portable Agent Skill (SKILL.md) that runs unchanged on Claude Code, Codex, and OpenCode; the agent-extension-author skill documents that cross-agent model. Each category has its own README with fuller blurbs.

tools — details

Skill	Description
forge-idea	Forge a rough idea into a viable one by fanning out parallel agents to stress-test and prune branches
llm-wiki	Build, grow, and query a linked Markdown knowledge wiki an LLM can navigate

model/tools — details

Skill	Description
prose-craft	Compose or revise clear, well-architected prose (Strunk & White + Pinker)
unslop	Edit prose to remove AI tells and restore a plain, human voice
agent-extension-author	Author, port, and harden skills and plugins across Claude Code, Codex, OpenCode, and oh-my-pi

engineering — details

Skill	Description
architect	Turn a fuzzy systems/infra requirement into a self-contained, agent-executable work package, then execute it

engineering/swe — details

Skill	Description
rust-engineer	Enforce Rust house guidelines while writing or reviewing Rust
code-quality	Audit code risk from complexity/coverage/duplication metrics + comment quality (reports only)

engineering/ci — details

Skill	Description
github-actions	Harden GitHub Actions workflows and configure Renovate

maintenance — details

Skill	Description
project-docs	Give any project a standard, enforceable docs/ (overview, spec, ADRs) with diagrams + a validator

environment — details

Skill	Description
nix	Work correctly in Nix + direnv dev repos — nix fmt, build traps, pre-commit tuning

Install

claude plugin marketplace add https://github.com/Teebor-Choka/skills

Or install a single skill:

claude plugin marketplace add https://github.com/Teebor-Choka/skills --plugin forge-idea

Repository layout

Skills are grouped into thematic categories for source organization. The grouping is invisible to the agent host — skills are discovered by their name field in SKILL.md.

tools/
  forge-idea/           # skill: forge-idea
  llm-wiki/             # skill: llm-wiki
model/tools/
  unslop/               # skill: unslop
  agent-extension-author/        # skill: agent-extension-author — references/platforms + validate_skill.py
engineering/
  architect/            # skill: architect — scaffold.py + lint.py, references, adapters
engineering/swe/
  rust-engineer/        # skill: rust-engineer
  code-quality/         # skill: code-quality — workflows: measure, comments
engineering/ci/
  github-actions/       # skill: github-actions — Actions hardening + Renovate
maintenance/
  project-docs/         # skill: project-docs — docs contract, validator, diagrams
environment/
  nix/                  # skill: nix — Nix + direnv dev workflow

Each skill is a self-contained directory:

<skill-name>/
├── SKILL.md        # Instructions + YAML frontmatter (name, description)
├── references/     # Reference files loaded on demand
└── assets/         # Copyable artifacts (templates, CI scripts, …)

Acknowledgements

architect

The architect skill was written from scratch, drawing on established architecture literature and document standards for their ideas, methodology, and a few short attributed quotes; it reproduces no substantial text from any of them, and references the one copyleft source (arc42, CC BY-SA) by concept only. With thanks to John Ousterhout (A Philosophy of Software Design); Mark Richards & Neal Ford (Fundamentals of Software Architecture, Building Evolutionary Architectures, Software Architecture: The Hard Parts); Martin Fowler (monolith-first); Michael Nygard (ADRs) and MADR; Alistair Mavin (EARS); arc42; and Simon Brown (C4 model).

Its graph-engineering step (references/graph-patterns.md) is grounded in the workflow / parallel / multi-agent-orchestration literature, every citation web-verified and authority-tagged, with the state of the art through 2026 folded in: the van der Aalst Workflow Patterns paper (2003), GPTSwarm (ICML 2024) and AFlow (ICLR 2025) on agents-as-optimizable-graphs, MaAS (ICML 2025) on query-dependent workflow-graph design, the topology-effects work of Shen et al. (EMNLP 2025) with G-Designer and AgentPrune/AGP, the MAST failure taxonomy (Cemri et al. 2025), and the Yue et al. 2026 survey of workflow optimization as agentic computation graphs. Full attribution with links, license notes, and preprint flags is in engineering/architect/references/sources.md and graph-patterns.md.

Grounding the current state can optionally run through a code graph, single- or multi-repo, via a code indexing MCP when available, with a graceful fall back to manual reading otherwise (references/code-graphing.md).

unslop

The unslop skill was not written from scratch. It was built by pooling the strongest open "unslop" (de-AI-writing) agent skills on GitHub, deduplicating their rules and word lists, and folding in the research they cite, then restructuring the result around a structure-first pass (surface word-swaps alone barely move AI detection). Full attribution with links is in model/tools/unslop/references/sources.md.

With thanks to the authors of the MIT-licensed skills we pooled from: theclaymethod/unslop, mshumer/unslop, MohamedAbdallah-14/unslop, asavvin-pixel/unslop, woerndl/unsloppify, mnapoli/skills, and the cursor/plugins pstack base (surfaced via ui-skills.com).

And to the research that shaped the structural pass: Wikipedia's Signs of AI writing (CC BY-SA 4.0, used as a factual reference), the StoryScope paper (arXiv:2604.03136), the UMD / Google DeepMind cliché-removal study, Berens & Kobak's excess-vocabulary analysis, and the Google developer style guide.

License

MIT — see LICENSE.