The open-source quiron skill, installable via the skills CLI for coding agents like Claude Code, Codex and Cursor, rewrites AI-generated prose to remove machine-writing tells and then measures the result against human writing baselines. It requires no API key and runs on the agent's existing model.
FeedHive published an open-source AI Slop Detector cataloging 120 writing patterns that make text read as AI-generated, organized into 9 categories with explanations, bad examples, and rewrites. It ships as an agent skill for Claude Code and Codex, a browser-based checker, and structured JSON for custom tooling. The tool targets style cleanup rather than authorship detection, which matters for practitioners trying to keep AI-assisted output readable.
FeedHive published an open-source AI Slop Detector cataloging 120 writing patterns that make text read as AI-generated, organized into 9 categories with explanations, bad examples, and rewrites. It ships as an agent skill for Claude Code and Codex, a browser-based checker, and structured JSON for custom tooling. The tool targets style cleanup rather than authorship detection, which matters for practitioners trying to keep AI-assisted output readable.
RUC DataLab released SkillAdam, an open-source tool (MIT license, Python 3.10+) that automatically improves AI agent SKILL.md files. It tests a skill on relevant tasks, learns from results and prior revision feedback, and applies only changes that pass evaluation. It ships as integrations for Codex, Claude Code, Cursor Agent, and GitHub Copilot, letting developers optimize skills inside their existing agent.
RUC DataLab released SkillAdam, an open-source tool (MIT license, Python 3.10+) that automatically improves AI agent SKILL.md files. It tests a skill on relevant tasks, learns from results and prior revision feedback, and applies only changes that pass evaluation. It ships as integrations for Codex, Claude Code, Cursor Agent, and GitHub Copilot, letting developers optimize skills inside their existing agent.
Unslop is a Claude skill that removes signs of AI writing from English text, including word-level tells like 'delve' and structural issues like over-explained conclusions. It calibrates to the user's personal voice, built on research from UMD and Google DeepMind.
Unslop is a Claude skill that removes signs of AI writing from English text, including word-level tells like 'delve' and structural issues like over-explained conclusions. It calibrates to the user's personal voice, built on research from UMD and Google DeepMind.
The sepia project published a portable Agent Skill that removes AI tells from writing across coding agents including Claude Code, Codex, Grok Build, and Antigravity. It repairs narrative architecture in fiction before word choice and applies venue-matched rules to professional documents. It is based on the StoryScope paper (arXiv:2604.03136).
The sepia project published a portable Agent Skill that removes AI tells from writing across coding agents including Claude Code, Codex, Grok Build, and Antigravity. It repairs narrative architecture in fiction before word choice and applies venue-matched rules to professional documents. It is based on the StoryScope paper (arXiv:2604.03136).
A developer released 'unflat', an Agent Skill for Claude Code, Cursor and Codex that fixes the flat 'AI mockup' look of agent-built sites. The agent audits the page, extracts design tokens, and appends one reversible CSS block adding a ground color, a light source, and mounted surfaces using the site's own colors. No markup changes are required.
A developer released 'unflat', an Agent Skill for Claude Code, Cursor and Codex that fixes the flat 'AI mockup' look of agent-built sites. The agent audits the page, extracts design tokens, and appends one reversible CSS block adding a ground color, a light source, and mounted surfaces using the site's own colors. No markup changes are required.