Contexory - Author agent skills with your team, sync to any AI agent
by•
Where teams author, review & version AI agent skills — then get them into every agent. Non-devs write them (no YAML); you keep git + review. Auto-syncs to Claude Code & OpenClaw; two-way GitHub sync feeds your own agents (LangGraph, custom). Portable markdown, no lock-in.
Replies
Hi PH 👋 I'm Desmond from Contexory.
Agent skills are the instructions that make an agent good at your work — and they're usually a mess: scattered SKILL.md files or repo configs, no version history, no review, copy-pasted between machines and teammates. The deeper problem: the person who should write a skill is whoever knows the workflow — ops, support, research — not whoever knows YAML. So they don't, and the skill ends up a developer's secondhand guess.
Contexory fixes both. Non-devs author in a real collaborative editor (an AI assistant handles the frontmatter/structure); you get versions + review/approval like code.
Then it gets the skill into your agent, two ways: • Claude Code & OpenClaw — published skills auto-sync via our plugins (on npm + ClawHub + the Claude Code marketplace). • Your own agent (LangGraph, CrewAI, custom) — two-way GitHub sync keeps skills as portable markdown checked into your repo, so your agent loads them the way it already does. No lock-in.
It's early and I'd genuinely love your hardest questions and "why not just a repo + PRs?" challenges (I think the non-dev authoring + review layer is the point — happy to argue it).
Would love a CLI flag or simple web UI for diffing two versions of a skill side by side before merging in updates from a teammate, since markdown changes can sneak in subtle prompt-level shifts that are tough to catch in a plain PR view.
@kardelenme36957 great idea!
Coming from a software engineering background, I baked a fair bit of git-inspired workflow into Contexory so a couple pieces of this already exist today.
Skills have version tracking with a side-by-side diff view, and GitHub↔Contexory syncs get a diff view for resolving conflicts, so you can see exactly what changed on each side before picking a winner.
Diffing is an area we're actively investing in — including bringing that same side-by-side view into the review workflow, so you can catch subtle prompt-level shifts before merging a teammate's update rather than after.
Thanks for the suggestion - this is exactly the type of feedback and validation we hoped launch day would surface!
Love how this lets non-devs actually author agent skills without touching YAML. One thing that would help us though — adding a visual diff viewer for skill changes alongside the markdown preview, so reviewers can see exactly what behavior shifted between versions rather than just the raw text diff.
Really glad you brought this up, @melisnceda1tw6. One thing that might already help: the diff isn't just raw text on our side. You can see it right in the rich-text editor the way you wrote it, with the changed words highlighted, or flip to the raw markdown view if you'd rather read it line by line. Whatever's easier on the eyes.
The part that really got us thinking is your framing around showing what behavior shifted rather than just the text. That's a genuinely different and harder problem: surfacing "this edit changes how the agent behaves" instead of "these three words changed." It's exactly the kind of thing we'd love to dig into to help non-dev authors thrive. Really appreciate the suggestion.