Skills overview¶

This page indexes the skills bundled in this repo.

Local skills (skills/local/) are authored and maintained here.
Owned skills (skills/owned/) are first-party skills maintained in another project and synced here for distribution; see the workflow.
Vendored skills (skills/vendor/) are cherry-picked from upstream repos and synced via the vendor.yaml manifest — see Adding vendor skills. Do not edit vendored SKILL.md files locally; changes will be clobbered on the next make sync.

Local skills¶

Custom-authored, curated to this repo's conventions (see Conventions).

Skill	One-line	Detailed page
`project-knowledge-harness`	TODO + backlog + pitfalls structure with a bundled validator/init/promote toolkit	docs
`quantatitive-factor-researcher`	Quantitative factor research persona for Python-based strategy work	docs
`vectorbt`	OSS/PRO library development, XDG documentation synchronization, and optional MCP	docs
`nautilus-trader`	v1 Cython / v2 Rust-PyO3 generation detection, version-matched docs cache, event-driven backtests, and live-trading gates	docs
`skill-author`	Author new skills following agentskills.io best practices; ships `new-skill.sh` and `lint-skill.sh`	docs
`verifiable-surfaces`	Design verifiable CLI/tool/service surfaces (`--help`/`--dry-run`/`--print-config`/isolated smoke) and verify config changes via app-native loaders	docs
`12-factor-agent-design-review`	Design or evidence-review production LLM applications across prompts, context, typed tools, durable state, owned control flow, pause/resume, humans, retries, and replay	docs
`openui-direct-provider`	Build OpenUI generative UI on your own model provider (no Gateway); provider-neutral companion to vendored `openui`	docs
`demo-evidence`	Capture acceptance evidence (screenshots/recordings/HTTP logs) into a gitignored `.evidence/` bundle keyed to git branch/commit + agent session, for async "Demos over diffs" review	docs
`mkdocs-site-bootstrap`	Bootstrap a MkDocs Material site + GitHub Pages deploy; consent-gated with `.skills/preferences.yaml`	docs
`zensical-site-bootstrap`	English Zensical pilot with safe initialization, native LLM exports, and build-only CI	docs
`marimo-batch-mlflow`	marimo dual-mode (UI + batch CLI) notebooks with Tyro + MLflow	docs
`dvc-ml-workflow`	DVC pipelines + queued experiments with metrics auto-bound to ephemeral commits; ships init/queue/lint helpers	docs
`mlflow-tracking`	Generic MLflow skill — sqlite + `mlflow ui`, vendored PostgreSQL + MinIO docker stack, LLM tracing, registry, autolog	docs
`agent-history-hygiene`	Commit transcripts/plans with feature diffs, derive staged harness/model provenance, bootstrap secret scanning, and apply rotate-first leak remediation	docs
`git-workflow`	Scale-aware Git workflow with English Conventional Commits, cross-harness AI provenance, linear history, worktrees, SemVer, and commit validation	docs
`superproject-workspace`	Keep a git-submodule superproject from drifting — per-child track/pin/lazy policy in `.gitmodules`, lag/drift status, preflighted one-commit bumps, scheduled bump PRs, child worktrees, audit/adopt	docs
`pueue-job-queue`	Drive Nukesor/pueue for queued/parallel/scheduled shell jobs; submit-one + DAG submitter + JSON-summary waiter; observed pueue 4.0.2 schema	docs
`slurm-hpc`	Portable Slurm know-how — sbatch skeleton, resource requests, job chaining (`--dependency=afterok` + the DependencyNeverSatisfied trap), and what actually fences GPU VRAM (shard vs MPS vs MIG)	docs
`long-running-jobs`	How an agent should wait for work outlasting a turn — scheduler-owned chaining, one blocking backgrounded wait, filtered event streams, scheduled check-ins last; durable exit-code markers via `run-and-mark.sh` + `check-runs.sh`	docs
`experiment-knowledge-harness`	Research memory for ML/Quant — LEDGER of overturnable findings, payoff-triaged ROADMAP, pre-registered REPORTs with single-axis ablation contract, auto-rendered Mermaid map	docs
`clash-proxy-api`	Discover & drive a Clash/mihomo external-controller: status/mode/TUN/switch/reload/connections + OS system-proxy toggle; multi-client (Verge Rev, ClashX, mihomo CLI) with enable-the-API guidance	docs
`fastapi-ai-patterns`	Production FastAPI patterns + gotchas for AI/ML/LLM serving; `def`/`async` decision table + 8 references over all 10 chapters	docs
`fastapi-ai-scaffold`	Generate a production-shaped FastAPI AI service (clean architecture, lifespan model, JWT, SSE, probes, tests, Docker); `new-fastapi-ai-service.sh` + 44-file skeleton	docs
`fastapi-ai-interview-prep`	100 self-written FastAPI/AI interview Q&A across 10 topics + a `quiz.py` mock-interview CLI	docs
`raycast-extension-dev`	Build/verify/ship Raycast extensions — the launchd PATH trap, the typecheck `ray build` skips, `MenuBarExtra` constraints, and the store checks `ray lint` never runs; ships a scaffolder + readiness checker	docs
`python-project-best-practice`	Modern Python project conventions — uv + src layout, Tyro CLIs, loguru, ruff/type/pytest behind a Justfile, an AGENTS.md docs-drift gate; ships a six-profile scaffolder and a 26-check read-only legacy audit	docs

Owned skills¶

Skill	Canonical project	Detailed page
`go-cli-tui`	awesome-lazy-tools	docs
`userscript-development`	Tampermonkey-Scripts	docs

Vendored skills¶

3rd-party skills cherry-picked because they fill a gap the local skills don't cover, or because the upstream is the canonical authority on the topic. The linked detail page shows what each skill teaches, upstream provenance, and last-sync commit is tracked in vendor.yaml.

Vendored skills can be flat (skills/vendor/<name>/) or grouped into a series (skills/vendor/<series>/<name>/). See Adding vendor skills for how series work.

Flat (notebooks + meta)¶

Skill	Upstream	Detailed page
`marimo-notebook`	marimo-team/skills	docs
`streamlit-to-marimo`	marimo-team/skills	docs
`anywidget`	marimo-team/skills	docs
`skill-creator`	anthropics/skills	docs
`herdr`	herdrdev/herdr	docs
`security-audit`	cloudflare/security-audit-skill	docs
`openui`	thesysdev/skills	docs

Fullstack Next.js series¶

series: fullstack-nextjs — Next.js (App Router) + Supabase (Postgres) + shadcn/ui + Tailwind CSS + design/testing skills. All from official orgs (Vercel, vercel-labs, Supabase, Anthropic).

Skill	Upstream	Detailed page
`nextjs`	vercel/vercel-plugin	docs
`shadcn`	vercel/vercel-plugin	docs
`react-best-practices`	vercel/vercel-plugin	docs
`vercel-storage`	vercel/vercel-plugin	docs
`supabase`	supabase/agent-skills	docs
`supabase-postgres-best-practices`	supabase/agent-skills	docs
`web-design-guidelines`	vercel-labs/agent-skills	docs
`frontend-design`	anthropics/skills	docs
`webapp-testing`	anthropics/skills	docs

For the rules every local skill follows (layout, naming, scripts, references), see Conventions. For how vendoring works end-to-end, see Adding vendor skills.