This is the case Contexory is built for.Experts edit in plain language, and a release reaches production only when you move the label to it.
Side by side
What mattersA shared docA doc the team edits and pastes fromA repo of prompt filesMarkdown in git, shipped by your buildContexoryFits your answerPlain-language editing with releases
Cost to startA shared docFree, and already openBest hereA repo of prompt filesFree if you already have a repoBest hereContexory · Fits your answer$9 a month for one person, $15 a seat for teams, after a 14-day trial
Who can edit comfortablyA shared docAnyoneBest hereA repo of prompt filesEngineers. Everyone else needs help with gitContexory · Fits your answerAnyone, in plain languageBest here
What people have to learnA shared docNothingBest hereA repo of prompt filesGit, Markdown and frontmatterContexory · Fits your answerA new tool, with no git or frontmatter
Editing togetherA shared docLive cursors and commentsBest hereA repo of prompt filesBranches and merges, one change at a timeContexory · Fits your answerReal-time co-editing with presence, on the Team planBest here
Getting it to the agentA shared docCopy and paste by handA repo of prompt filesYour own build or sync scriptContexory · Fits your answerClaude Code and OpenClaw plugins, GitHub sync and a signed APIBest here
HistoryA shared docEdit history, no numbered releasesA repo of prompt filesFull history and branchesBest hereContexory · Fits your answerVersion history, plus numbered releases that never changeBest here
Review before it shipsA shared docCommentsA repo of prompt filesPull requestsBest hereContexory · Fits your answerSkill reviews and approvals, on the Team plan
Checks before go-liveA shared docNoneA repo of prompt filesWhatever CI you writeBest hereContexory · Fits your answerCases your team writes. A failing release is held out of production unless an admin overrides it with a reasonBest here
RollbackA shared docFind an old version and paste it backA repo of prompt filesRevert and redeployContexory · Fits your answerMove the production label back to an earlier releaseBest here
Packaging for each hostA shared docUp to youA repo of prompt filesUp to youContexory · Fits your answerThe plugins lay out each host's files from one set of contextBest here
Tool permissionsA shared docNot coveredA repo of prompt filesConfig files you maintainContexory · Fits your answerPer-tool limits on each agent, named in plain words and marked reads or writes. A tool Contexory cannot classify stays offBest here
Lock-inA shared docNone. It is your docBest hereA repo of prompt filesNone. It is your repoBest hereContexory · Fits your answerLow. Skills are plain Markdown, and GitHub sync keeps a copy in your repo
“Best here” marks the strongest option on that row. Where two are equal, both are marked.
When to keep what you have
Keep the shared doc if
one person writes the prompt and pastes it in by hand
the agent is an experiment nobody outside the team uses
a bad edit costs you a redo, and nothing more
Teams usually move on when a second editor joins or the agent starts answering customers.
Keep the repo if
everyone who edits the context is comfortable with git
you already run the checks you need in CI
pull requests are how your team prefers to review
Teams usually move on when the people who know the answers are not the people who can open a pull request.
Switching
Moving in is a few steps. Moving out is a copy of plain files.
From a repo
Connect GitHub and import
Connect the repository, and Contexory finds the folders that hold a SKILL.md and shows what it would bring in before anything is created. Importing an agent also reads its AGENTS.md and CLAUDE.md, and never writes to the repository.
AGENTS.md
skills/<name>/SKILL.md
skills/<name>/references/
From a doc
Paste it into a new skill
Paste the playbook you run today. The assistant drafts the skill and asks about the gaps it finds, and nothing is created until you approve it. It reaches an agent only once you publish it.
Leaving
Take your files with you
Skills are SKILL.md folders and plain Markdown. GitHub sync writes them to your own repo, the plugins keep a copy on the machine where your agent runs, and an API token can download any skill as a .tar.gz. There is nothing to convert.
SKILL.md
references/
scripts/
Try it with one skill you already have.
Paste a playbook or connect the repo, then compare for yourself. If the doc or the repo still suits you better, you keep the Markdown.