groove-utilities-prime Outcome The agent receives full groove workflow context in the conversation — config, key commands, conventions, and constraints. Run at the start of every session before doing any work. Acceptance Criteria Agent is shown current config values from .groove/index.md Agent is shown all key commands, conventions, and constraints If a newer version of groove is available, agent is notified Output format Output the following to the conversation (do not write to any file):
Groove Workflow Context
groove is installed in this repo. Use /groove-* skills for all workflow commands.
Config
tasks:
Key commands
/groove-daily-start — start the workday /groove-daily-end — end the workday /groove-work-brainstorm — clarify scope (YAGNI enforced) /groove-work-plan — research codebase, write implementation plan /groove-work-exec — execute the plan /groove-work-review — evaluate output, decide accept/rework /groove-work-compound — document lessons into existing project files /groove-work-spec — create outcome spec (what to build) /groove-work-doc — create reference doc (how it works) /groove-utilities-task-list — show active, ready tasks /groove-utilities-task-create — create a task /groove-utilities-task-analyse — summarise tasks by status /groove-utilities-memory-log-daily — write daily end log /groove-utilities-memory-promises — capture or resolve deferred items /groove-utilities-memory-mistakes — log a mistake and resolve it /groove-utilities-memory-retrospective — analyse ratings, mistakes, and learnings over a period /groove-utilities-memory-graduate — promote a stable lesson to AGENTS.md permanently
Conventions
- Stage tasks: "YYYY-MM-DD,
" (no numbers; e.g. 2026-02-28, Brainstorm; 2026-02-28, Compound — topic) - Memory logs:
daily/, weekly/, monthly/, git/ - Task completion requires "Summary of Changes" in body before marking done
- Archive is always user-triggered — never automatic during end
- 80% of compound loop value is in plan and review — do not skip them
Steering
- Fix root causes, not symptoms — if a workaround is needed, note the root cause anyway
- YAGNI — only make changes directly requested or clearly necessary; no unrequested refactors
- Verify before acting on shared state — confirm before push, PR creation, or destructive ops
- Smallest diff that solves the problem — prefer editing one file over touching five
- When blocked, ask rather than brute-force — retrying the same failing action wastes context
- Read before editing — understand existing code before proposing changes to it
Constraints
- Do not edit files under
skills/or.agents/skills/— managed by groove update, changes will be overwritten - User zone:
.groove/is yours — config, hooks, memory, and learned insights are all safe to edit If groovebook: is set in .groove/index.md , append to the output:
Groovebook
groovebook:
Identity
section header If it does not exist: skip silently — no prompt, no hint The identity section is free-form — output the file contents verbatim. This gives the agent persistent user context (mission, goals, projects, beliefs) across sessions without re-explaining each time. Constraints Read .groove/index.md frontmatter to substitute tasks: , memory: , and git.* placeholders Output to conversation only — do not write to AGENTS.md or any other file (except updating .groove/.cache/last-version-check ) Version check runs at most once per day — gate on .groove/.cache/last-version-check date