惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
IT之家
IT之家
博客园_首页
博客园 - 【当耐特】
V
V2EX
Apple Machine Learning Research
Apple Machine Learning Research
G
Google Developers Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Recent Announcements
Recent Announcements
F
Fortinet All Blogs
GbyAI
GbyAI
腾讯CDC
H
Hackread – Cybersecurity News, Data Breaches, AI and More
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
I
InfoQ
H
Help Net Security
T
Tailwind CSS Blog
B
Blog RSS Feed
Martin Fowler
Martin Fowler
人人都是产品经理
人人都是产品经理
The Cloudflare Blog
博客园 - 叶小钗
雷峰网
雷峰网
量子位

Show HN

GitHub - astefanutti/shaderbang: Shebang for Shaders Show HN: AI agents for UK GDAD PCF roles and their skills The Two Pillars: Mixer Mode and Meta-Software in the Reorganization of Software Work After AI GitHub - JaiCode08/teleport-env What 1,000+ Harness Experiments Taught Me About Self-Improving Agents Show HN: Liiists, a Markdown-first, iOS and CLI list app SwiperTab – Get this Extension for 🦊 Firefox (en-US) GitHub - kouhxp/fftext: Summarize, explain, fact-check, or translate any text, URL, or file. No GPU. No cloud. One command GitHub - sweetpad-dev/sweetpad: Develop Swift/iOS projects using VSCode GitHub - dogmaticdev/IRON: IRON a.k.a. Intermediate Representation Object Notation is a Interpreter/Database that is used to create Programming Languages. GitHub - sjhalani7/vaen: Package your AI coding harness into a portable .agent file, and share it across repos, teams, & the community without ever having to copy-paste instructions, skills, MCP config, or secrets. Show HN: Gandalf the Grader Show HN: Citadeld – replay any CI failure locally from a single file GitHub - tdortman/cuSBF: High-Performance GPU Super Bloom Filter coral-ai/claude-code-token-xray at main · Coral-Bricks-AI/coral-ai GitHub - ulyssestenn/funes: Funes is a Git-based framework for LLM-managed knowledge work: an AI Librarian ingests raw sources, builds an interlinked Markdown knowledge base, and uses it to produce cited reports, analyses, and other outputs. GitHub - ThatXliner/gah: Git Add Hunk, built for agents to use GitHub - harmont-dev/harmont-cli: Command-line client for the Harmont CI platform GitHub - brooksmcmillin/mcp-authflow: OAuth 2.0 Authorization Server framework for MCP servers GitHub - javaid-codes/audit-supply-chain-agents GitHub - amorey/gochan: A small library of common channel architectures for Go, inspired by Rust GitHub - arifozgun/OpenGem: Free, Open-Source AI API Gateway with Gemini, OpenAI & Anthropic Compatibility in 1 file GitHub - Pranesh950/BioPetals: 🌸 Run BIOxAI models at home, BitTorrent-style. Fine-tuning and inference up to 10x faster than offloading GitHub - cnguyen14/bounty-doctor: Diagnose a GitHub bounty issue before you waste hours: detects honeypot scam repos, AI-bot attempt swarms, and stale contests. Show HN: CoreMCP – MCP Server for On-Prem DBs Show HN: KittyHTML – Render HTML/CSS as an inline image in your terminal GitHub - bingud/filemat: Web-based file manager Show HN: TruthLens – Free multi-signal deepfake image detector GitHub - apexlocal-jz/claude-usage-tray: Windows system-tray app showing your Claude Code rate-limit usage at a glance. Zero deps, ~300 lines of PowerShell. Cross-IDE (works regardless of VS Code, Cursor, plain terminal). Release v0.1.2.1 · kouhxp/yapsnap
Your AI's memory file is rotting. Here's a free skill to ...
Pure Mint Software · 2026-06-26 · via Show HN

I built a skill for Claude because the memories it was saving about me had started to bog it down instead of help. The file kept growing, and the bigger it got the worse Claude got with it.

The memory file is just where Claude writes down the rules, preferences and little facts it learns from you, so it can sculpt itself towards the way you actually want it to work.

Why memory files always bloat

The trouble is the file only ever grows in one direction. You keep adding, and it bloats. And the more you stuff into a large language model's context, the worse it gets, not only at recalling any single fact, but at staying consistent. Before long you've filled it with contradictions, rules quietly fighting each other, instead of a clean set of preferences steering the model the way you meant.

A skill that prunes the file with you, one line at a time

So I made the skill interview you, rather than deciding on its own what to keep and what to kill. It works through the file line by line and points things out: this phrase is repeated, cut it and I'll still understand the shorter version, so you save context; or, this line contradicts something we read three sentences ago, so which one do you actually want me to follow? It goes through the whole file and tidies up the crap, but only ever in a way you've agreed to.

It won't quietly make its own decisions and leave you with a set of hollowed-out preferences that no longer steer it properly. At every step it asks what matters, so nothing important gets cut by accident. And it shows you a diff: the original line on top, the suggested version underneath, so you can see exactly which words are going and which are staying. You agree, you reword it, or you tell it to leave that memory alone.

Download the Memory Cleanup skill (free) →

Unzip it, drop the folder with the SKILL.md in alongside your other Claude skills (ask Claude where that lives if you're not sure), and say "let's do a memory cleanup". I run it every week or two. A lean memory file is one the assistant actually follows.


I made it free because I wanted it for myself, couldn't find it anywhere, and it doesn't feel right charging for a text file that does a job, even though I know plenty of people would.

I also make software. Some of it's free, some of it's a small one-off payment. No bullshit: no subscriptions, no telemetry, none of that rubbish. Just simple software for people who remember how software used to behave back in the 80s, 90s and early 2000s before Facebook and MicroSlop and the general corporate machine started ruining it. John Carmack and Chris Sawyer are my coding heroes, not the bastards who invented popups and dark patterns.