Catnap
GNU-style sleep that reports time remaining on stderr while stdout stays clean. Scripts read silence; humans read progress.
Documentation
Guides generated straight from each project’s repository, grouped by what the project is. Every card links the source, the latest release, and the docs.
Command-line tools that do one thing well and then get out of the way.
GNU-style sleep that reports time remaining on stderr while stdout stays clean. Scripts read silence; humans read progress.
Topic-based FIFO task queue for Claude Code. Queue work with q put, dequeue when context allows – nothing gets forgotten between sessions.
Enforces consistent GitHub repository settings across an estate of enrolled repos by driving OpenTofu. Plan, review, apply – settings drift ends here.
Creates and updates GitHub labels from a TOML file or a sensible default set. Idempotent, Enterprise-aware, and deletes nothing.
Gives agents a view of computed styles. Captures deduplicated CSS snapshots from a live page via agent-browser, the Chromium DevTools Protocol, or Playwright – DevTools for terminals.
Code review with a hard per-review cost ceiling. Bounded model lanes, admission control, and a dakar-review CLI.
Linter for Differential Datalog. A Logos tokenizer, an error-recovering Chumsky parser, and a Rowan syntax tree – lints that survive broken input.
MCP server giving coding agents persistent semantic memory. Qdrant vectors and local FastEmbed embeddings – no cloud round-trips.
ODW workflows plus a supervisor skill that drive a project from roadmap through planning, review, implementation, and merge. Happy robots build happy code.
Shared lint tooling for JavaScript and TypeScript: an Oxlint plugin with local rules for conditional complexity and documentation.
Local DigitalOcean v2 API simulator. A deterministic, seeded control plane with action polling and eventual consistency for infrastructure tests.
Terminal code review assistant. Pulls pull-request comments into a keyboard-driven TUI with SQLite caching and background refresh.
Git subcommands that make branch-based development less of a… well, you know. Linked worktrees, branch tracking, and repository scaffolding.
Architecture checker for hexagonal Python projects. Walks package roots with ast, classifies imports, and reports boundary violations.
Coordinates version bumps and publication plans across a Rust workspace from one lading.toml. Dry runs before releases.
Splices one Markdown roadmap into another. Structural edits, renumbered phases, and rewritten cross-references – no manual renumbering.
Unb0rks Markdown tables. Reflows columns to uniform widths and wraps prose to 80 columns with Markdown-aware tokenization.
Syncs the working directory to a short-lived cloud VM, runs the build there over SSH, streams output back, and destroys the evidence.
YAML-powered build system that fuses the minimalism of Ninja with the templating nous of Jinja. Write once, build anywhere—reproducible, declarative, and just opinionated enough to keep you honest.
Finds Mermaid blocks in Markdown and renders every one to prove it compiles. Broken diagrams fail the build, not the reader.
Static linter for Open Dynamic Workflow files. Checks workflows before they run – without executing source or dispatching agents.
Runs AI coding agents inside isolated Podman containers. Credentials and the container socket stay with the trusted host side.
Temu-tier isolation for unit testing. Pulls container images, exports them to disposable root filesystems, and executes commands via podman, bwrap, proot, or chroot.
Claude Code stop hook that works out which files a turn changed, runs the matching quality targets, and blocks the turn until they pass.
Renders shared MCP configuration and reusable agent skills from one XDG data directory into repository-local files for Claude, Codex, and friends.
Installs, version-checks, and runs the Kani and Verus verification toolchains for Rust repositories. One CLI instead of four READMEs.
Prose linter for Markdown and source-code doc comments. Tree-sitter extraction, a Rust core, and Python bindings – style enforcement that reads the code too.
Generates en-GB-oxendict typos.toml configuration. Refreshes a shared Oxford dictionary cache and merges repository-local overrides.
Fetches unresolved pull-request review comments and renders them as terminal Markdown. Issues and CI status included.
Semantic CLI for codebases. A light client, a weaverd daemon, sandboxed language servers, and tree-sitter – observe, act, verify.
Nine opinionated Dylint lints for Rust, plus one experimental opt-in. Snark included at no extra charge.
Crates and packages built to be depended on: typed, tested, and documented.
Shared Actix Web building blocks: keyset cursor pagination, idempotency handling, server-sent events, and OpenAPI helpers.
Work-in-progress FISHDBC engine written in Rust that merges a rayon-driven CPU path, Rust-CUDA kernels, and a C-ABI plugin system. Built for safe, iterative clustering across heterogeneous hardware—sushi-grade data science.
Python-native command mocking harness that intercepts external processes over Unix domain sockets so pytest suites can spy, stub, and replay shell workflows without brittle bash fixtures.
Typed, async-first process execution from a curated catalogue of approved executables. Structured results, synchronous wrappers included.
Common Table Expressions for Diesel. Builders for recursive and non-recursive CTEs, with column metadata helpers.
Correlation IDs for Falcon web applications. One middleware, traceable requests.
Falcon extension for structured WebSocket routing and event orchestration. Define msgspec payload schemas, spin up background workers, and broadcast updates to rooms—all without turning your ASGI app into spaghetti.
Hybrid Python/Rust logging library that channels picologging’s minimalism through PyO3. Log records stream through crossbeam-channel workers for thread-friendly, low-latency structured logging.
Low-dependency Rust microcrate for injecting monotonic clocks. Elapsed-time logic becomes deterministic under test.
Configuration engine inspired by esbuild’s ergonomics—one schema, many surfaces. Generate matching CLI flags, env vars, and config files for Rust services without losing type safety or your mind.
Bootstrap postgresql_embedded clusters as root while delegating filesystem work to unprivileged users. Keep CI environments reproducible without burning down your permissions tree.
Pylint on PyPy without the crashes. Patches Astroid's runtime introspection so linting works where CPython assumptions don't hold.
Behavior-driven testing macros layered on top of rstest. Narrative clarity without the ceremony.
pytest fixtures that run tests against a local, stateful GitHub API simulator. Scenario configuration, seeded repository state, and no rate limits.
The stateful mock GitHub REST and GraphQL engine behind Simulacat, built on Simulacrum.
Record/replay HTTP proxy for OpenAI-compatible chat APIs. Captures real exchanges into cassettes for deterministic tests.
Podcast-grade TEI P5: a Rust data model with parser and emitter, exposed to Python via PyO3 and msgspec.
Formally verified theorems as engineering artefacts. A theorem_file! macro expands .theorem schemas into compile-time verification modules for Kani.
The recommendation engine behind Wildside. Domain model, OSM and Wikidata ETL, and relevance scoring as a Cargo workspace.
Framework for building binary protocol servers and clients in Rust—think Actix Web, but for everything not HTTP. Message routing, framing, and backpressure included; bring your own transport.
Prompt-injection defence for tool-using agents. Capability-based control- and data-flow policy enforced at runtime, following the CaMeL design.
Applications for desktop and browser. Products with a surface, not just a prompt.
Local-first personal AI assistant. REPL, web gateway, webhooks, routines, and background jobs with sandboxed WASM tools.
AI agent orchestration that keeps canonical conversation and task history across backend handoffs. The best room, not the smartest agent.
Agentic podcast studio. Turns heterogeneous sources into scripted, mastered audio with a human in the loop for polish.
Vector illustration with pervasive accessibility and scriptability. A GPUI editor with SVG import and export, plus undo history that remembers selections.
Estate-level status reporting. Ingests GitHub activity through a medallion pipeline and produces repository health reports.
Hotline-style daemon for joyful file-sharing and retro chat. Tokio, Diesel, and Argon2 behind a protocol from 1997.
Self-hosted appliance that indexes private repositories into a semantic search corpus. grepai and Qdrant, without leaving the building.
Modern, extensible terminal built on open web standards. A Hyper fork with a command-driven architecture, WebGL rendering, and a robust plugin system.
Serendipity for urban wanderers. Narrative walking routes from OpenStreetMap and Wikidata, shipped as a PWA with a Rust backend.
Playful worlds in executable form.
Overengineered real-time strategy. A dataflow-driven game loop built on DBSP, with optional Bevy rendering subsystems.
Offline-first idle strategy about shipping software. Allocate effort across shipping, open source, quality, security, and marketing; a deterministic simulation decides what breaks.
Skill collections that teach coding agents the house style.
Bootstrap scripts, hooks, sub-agents, and a bundled skill collection for provisioning agent development environments across the estate.
Claude Code skill that writes novels under a Ralph Loop harness – treatment, beats, outline, draft, revision, repeat until done.
Compact Python skill catalogue for coding agents: one router, focused language skills, a verification selector, and a quality-tools guide.
Compact Rust skill catalogue for coding agents. Keeps the useful decision surfaces, drops the boilerplate.
Agent skills for AI film production – from prose to picture. Continuity extraction, reference images, shot direction, and model-routed video prompts, with a TTS phoneticizer for good measure.