Small networking and hardware tools, a Windows memory forensics monitor, and a shed for the scripts that belong to none of them. Mostly Python, mostly no dependencies, and each one built to be read as well as run.

Project	What it is	Latest	Build
nettail	tail -f for your network: a NetFlow and IPFIX collector that prints flows with hostnames, colour and a live status bar
netflume	NetFlow v5, NetFlow v9 and IPFIX collection and parsing, as a library
lanname	Address to hostname lookup on a local network: reverse DNS, mDNS and NetBIOS, cached and non-blocking
readerboard	An HTTP service for BetaBrite and Alpha protocol LED signs, with alerts, scheduling and clock sync
bravia-http-remote	A single-page console for controlling a Sony Bravia display over your network
cec-ir-bridge	Volume control for an IR-only soundbar from the Apple TV Remote app, over HDMI-CEC and an ESP32 IR blaster
toolshed	Small scripts and helper apps that belong to no particular project, one directory per tool, and mixed languages on purpose	no releases
Memlapse	A Windows memory forensics monitor: live process and memory-map views, recording and timeline replay of a process's memory, and heuristics for in-memory code injection	no releases

The badges above say whether the last build passed and what the registries serve. They cannot say whether those two agree with each other, or whether a tag is on the branch it claims to be on, so the table below asks, of every repository in the table above.

scripts/sweep.py reads it fresh from GitHub, PyPI and both container registries every morning and writes what it found here. It is standard library only and clones nothing. When something is genuinely wrong, as opposed to merely untidy, the run fails rather than quietly updating a table nobody reads.

_{Generated: 2026-09-10 12:06 UTC}

Read the columns this way. Tags on main counts tags that are ancestors of main; a tag put on a release branch before it was squash merged is not one, and the history stops describing it. Unreleased is how far main has moved since the newest release, which is normal in itself. Published is what the registries actually serve, asked of them rather than inferred from a green publish job. CI is the latest run of each workflow on main, leaving out any the sweep is told is a fixture rather than a check: toolshed's Dungeon Crawl exists to be dispatched at and fails on purpose when fed bad input, so its runs are not read. A repository that releases nothing, which is what the shed is and what Memlapse is for now, shows none and n/a where those questions do not apply, and the columns that say anything about it are CI, PRs and Unmerged.

Nothing third-party is used within seven days of its release. Dependabot has a cooldown setting for that, and the repositories above that use Dependabot set it, but it only holds where the registry tells Dependabot when a version came out, and Docker Hub does not. scripts/are_you_cool_enough.py asks the registries itself, every morning after the sweep, for each open Dependabot pull request in the repositories above, and writes the ones that have waited long enough here, then the ones still waiting with the moment each clears. One it could not date at all is only in the run's summary, with the reason. It reads the project table above for the repositories to check, so a row added there is a row checked here.

Nothing yet.

_{Checked: 2026-09-09 12:25 UTC}

Claude Code is my preferred coding assistant, and this is the one place it is named. Individual projects, commits and files do not credit the agent or model behind them; a credit on every item reads as an advertisement. The same thinking puts agent instructions in AGENTS.md rather than one vendor's file, and keeps JetBrains and VS Code configurations in parallel. I am not in the business of promoting one commercial tool over another, and none of them should be a condition of using or adopting anything here.