drom-flow

Structured workflows, parallel grok sub-agents, and closed-loop pipelines for Claude Code

View the Project on GitHub drompincen/drom-flow

Install, update, uninstall

Why scripts ship as text

GitHub routes ZIP downloads through codeload.github.com when a repo contains .sh files, which some corporate firewalls block. So *.sh is gitignored and every script’s full source lives in SCRIPTS.md. Generate them once:

claude "Read start-here.md and follow the setup instructions"

SCRIPTS.md is the real distribution channel — a script change that isn’t re-embedded there never reaches anyone.

Install / update / uninstall

Your skills are yours

drom-flow ships skills, and teams tune them. On update, a skill file you have edited is kept — the new version lands beside it as <name>.dromflow-new and the run tells you which files it preserved. A skill you have not touched updates normally, so the policy preserves without freezing. Skills drom-flow does not ship are never read, moved or deleted, and --uninstall leaves both your own skills and any you edited in place.

This applies to .claude/skills/ only. Hooks, scripts and the engine still update in place with a backup, because freezing those on a local edit would silently deny you every future fix.

Shell assets ship as text (*.sh.txt) so ZIP downloads and corporate mail/proxy filters cannot strip them; init materialises runnable *.sh on install and update.

bash /path/to/drom-flow/init.sh.txt              # install into the current project
bash /path/to/drom-flow/init.sh.txt --check .    # dry run
bash /path/to/drom-flow/init.sh.txt --update .   # upgrade
bash /path/to/drom-flow/init.sh.txt --uninstall .

Never overwritten: CLAUDE.md, context/MEMORY.md, context/DECISIONS.md, context/CONVENTIONS.md, scripts/orchestrate.sh, plans, reports. Overwritten files are backed up to setup-backup/<timestamp>/ first.

--update also merges new guidance sections into an existing CLAUDE.md rather than replacing it, and preserves third-party hooks in .claude/settings.json (a tool like hyperresearch registering its own hook is not clobbered).

What a host project receives

Path Contents
CLAUDE.md behavioural rules, parallelism, closed-loop and plan protocol
.claude/skills/ the skills marked ships in Skills
.claude/hooks/ lifecycle hooks (see Hooks)
.claude/docs/ operator runbooks — gitignored, including runbook.md
scripts/ the fleet, research, audit and watcher scripts
workflows/, context/, drom-plans/, reports/ protocol, memory, plans, run output

This site is never installed. docs/ is the public guide and stays in the repo; host projects get .claude/docs/ instead — terse, operational, and gitignored so it never pollutes your tree.