GitHub Projects Community

Explore it here: opensourceprojects.dev/post/pilotdeck

PilotDeck is an open-source agent OS with white-box, traceable WorkSpaces for long-running, multi-project productivity. - Memory that is white-box and traceable, editable per entry - Token cost tracked per task for economic viability - Automatic model routing based on task difficulty - Proactive agent execution that reports progress and lands files on disk

Explore it here: opensourceprojects.dev/post/codexguide

CodexGuide maps how to use Codex as an AI workflow system across CLI, cloud, IDE, desktop, and mobile, not just as a code tool. - Covers beginners through teams: setup, subscriptions, first tasks, and real project integration - Documents CLI config, AGENTS.md, sandbox approvals, and team playbook for production use - Includes practical recipes for PPT, Obsidian, Figma, CI fixes, and clinical literature reviews - Maintains a verified index of OpenAI official docs and sources with last-checked dates

Explore it here: opensourceprojects.dev/post/skillopt

SkillOpt trains agent skills using neural network training concepts (epochs, batch size, learning rate) without modifying model weights. - Supports training on SearchQA, ALFWorld, DocVQA, and other benchmarks - Uses optimizer and target model separation for skill refinement - Validates skills through training/validation/test splits - Works with Azure OpenAI, OpenAI, Anthropic Claude, and local Qwen models

Hey @grok, what are they cooking? 👀

Explore it here: opensourceprojects.dev/post/ktx-ai-da…

ktx is an open-source context layer that automatically builds and maintains the semantic and business knowledge agents need to query your warehouse accurately. - Ingests wiki content, removes duplicates, and flags contradictions for human review - Detects joinable columns and resolves chasm and fan traps automatically - Combines full-text and semantic search across wiki and semantic-layer entities - Ships CLI and MCP tools for agent execution

Kimi Code CLI is a terminal-based AI agent that edits code, runs shell commands, and fetches web pages, designed for fast, focused coding sessions. - Single-binary install with no Node.js required - Purpose-built TUI for long agent sessions - Video input support for screen recordings and demo clips - Subagents for parallel, isolated work (coder, explore, plan) Explore it here: osp.fyi/kimi-code

Explore it here: opensourceprojects.dev/post/hexyl

hexyl is a terminal hex viewer that uses color to categorize bytes by type - NULL, printable ASCII, whitespace, other ASCII, and non-ASCII. - Color-coded output distinguishes byte categories at a glance. - Available via apt, brew, cargo, snap, and many other package managers. - Supports custom color schemes through environment variables. - Works on Linux, macOS, Windows, and BSD systems.

GitHub Repo: github.com/cosmicstack-la…

What npm packages did for apps, skills may do for AI agents. Mercury Skills is building an open ecosystem of reusable SKILL•md workflows for Mercury, Claude Code, Codex CLI, Cursor, Gemini CLI, and more. One skill. Multiple agents.

GitHub

@TTrimoreau GitHub

Name a tech company that nobody hates

Explore it here: opensourceprojects.dev/post/pullfrog

Pullfrog uses GitHub Actions to connect your coding agents directly to GitHub events like issues, PRs, and reviews. - Tag `@pullfrog` in comments to trigger agent tasks with full repo context - Configure automated agent runs for PR creation, issue labeling, and more - Agents can auto-fix CI failures and address PR reviews - Works with any coding agent via a built-in MCP server

Happens all the time

Explore it here: osp.fyi/stop-slop

Stop Slop is a skill for Claude and other LLMs that teaches them to remove the predictable phrases and structures that make AI-generated text sound robotic. - Catches banned phrases like throat-clearing openers, emphasis crutches, and all adverbs - Eliminates structural clichés such as binary contrasts and dramatic fragmentation - Enforces sentence-level rules like no Wh- starters and active voice only - Includes a 5-dimension scoring rubric to evaluate directness, rhythm, and authenticity