PromptOps
Manage agent instructions like feature flags. Enable or disable specific prompts without touching the codebase, keep org policy and private prompts separate, and only turn on instructions your setup can actually honor.
Prompts as composable fragments
Your CLAUDE.md, roles, and playbooks are broken into addressable fragments, each individually toggleable. At session start, the SessionStart hook composes your effective instruction set from whatever is enabled — so flipping a prompt off in the dashboard takes effect on your next session with no code edit.
Org vs individual
Org-shared fragments are managed by an org admin and apply to everyone on the team. Individual fragments are private to you — they live in a gitignored path and never hit the repo, but our tooling re-injects them at session start. Org policy fragments can be locked: an individual can extend them but never silently weaken or disable them.
The wire-up dependency chain
A prompt only works if its prerequisites exist. Each prompt declares the capabilities it needs; as you wire more in, more prompts light up. The dashboard shows this live for your account — below is the canonical chain.
The CI merge-gate tradeoff
The sprint merge gate checks a PR against its approved spec. Whether it can block a merge — versus just advise — depends on your GitHub plan and repo visibility:
| Setup | Merge gate | Why |
|---|---|---|
| Public repo | enforceable | Required status checks can block merges on public repos. |
| Paid plan (private repo) | enforceable | Paid plans get required-check enforcement on private repos. |
| Free plan + private repo | advisory | GitHub can't enforce required checks here — the gate nudges but never blocks. |
Progressive enablement
You don't have to wire everything up front. Start with what you have; as each integration comes online, the dashboard offers to enable the prompts it just unlocked. Open the live view at the Prompts dashboard to see your current readiness and toggle fragments.