A Claude Code agent skill that teaches an agent
to share markdown for async review and pull the feedback back — using the
dm CLI.
It's the highest-level way to use Draftmark from an agent: instead of teaching
the model the REST API or a list of curl commands, you hand it one capability
and it knows the whole loop — publish a plan, share the link, read comments,
iterate, close the review.
The skill wraps the dm CLI and encodes the workflow:
- Publish —
dm create plan.md --agent→ share URL - Hand off — surface the link for humans to review
- Read feedback —
dm comments --json,dm status --json - Iterate & close —
dm update,dm close
It also covers acting as an agent reviewer, feedback-gate exit codes, and
cross-session continuity via .draftmark.json.
- The
dmCLI:npm install -g draftmark - A Draftmark account key (
acct_...) for private docs / cross-session reads:dm login --api-key acct_xxxxorexport DM_API_KEY=acct_xxxx(public docs work without a key)
Clone the repo and copy (or symlink) the draftmark/ folder into your Claude
Code skills directory:
git clone https://github.com/draftmark-app/skill.git draftmark-skill
# user-level (all projects)
cp -r draftmark-skill/draftmark ~/.claude/skills/draftmark
# or project-level
cp -r draftmark-skill/draftmark /path/to/project/.claude/skills/draftmarkThen start (or restart) Claude Code — the skill is discovered automatically and activates on prompts like "share this plan for review" or "any comments on the doc yet?".
dm whoami # confirms the CLI is installed and authenticated- Draftmark: https://draftmark.app
- CLI on npm: https://www.npmjs.com/package/draftmark
- CLI source: https://github.com/draftmark-app/cli
- API docs: https://draftmark.app/docs
MIT