groove-utilities-stats — Claude Code skill: what it does
groove-utilities-stats prints a read-only dashboard over the groove memory folder: lessons captured and graduated, daily-log streaks, task counts and commits for a chosen period.
MIT 265 installs ★ 6 updated 2026-07-24
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a claude-codeWhat groove-utilities-stats does
The skill reads .groove/memory/ and prints a dashboard into the conversation. Three sections. Compound funnel: lessons logged, lessons marked as graduated, and the graduation rate. Adherence: daily-log coverage across the last memory.review_days business days, the current streak, whether weekly and monthly rollups are fresh, plus count and average of ratings taken from signals.md. Velocity: open and closed tasks and the number of commits in the period.
Pass week, month or all to pick the window. A bash script at scripts/stats.sh does the whole job without calling the model. Exit 0 means the output is ready to report. Exit 3 means local metrics printed but task counts from linear or github still need the markdown steps.
Nothing gets written, so the run is safe on a dirty working tree. The catch: everything sits under .groove/memory/, and a repo without .groove/ has nothing to count — the script exits 1 and points at the install step. Task numbers also depend on tasks.storage in .groove/index.md naming a backend; with none you get commits only.
What it helps with
- see how many lessons were logged and how many graduated
- catch a broken daily-log streak
- count open and closed tasks for a period
- check whether weekly and monthly rollups went stale
- average the day ratings from
signals.md
Example prompts for the agent
- Show the groove stats dashboard for
month - How many lessons did we capture and graduate this week?
- Print stats for
allwith task counts and commit velocity
How to install groove-utilities-stats
1Skills CLI (Vercel)
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a claude-code2Manually: put the files in place
~/.claude/skills/groove-utilities-stats/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/groove-utilities-stats/
1Skills CLI (Vercel)
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a codex2Manually: put the files in place
.agents/skills/groove-utilities-stats/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a cursor2Manually: put the files in place
.cursor/skills/groove-utilities-stats/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install andreadellacorte/groove skills/groove-utilities-stats2Manually: put the files in place
.github/skills/groove-utilities-stats/SKILL.md1Skills CLI (Vercel)
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a gemini-cli2Platform CLI
gemini skills install https://github.com/andreadellacorte/groove --path skills/groove-utilities-statsManually: put the files in place
.devin/skills/groove-utilities-stats/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add andreadellacorte/groove --skill groove-utilities-stats -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about groove-utilities-stats
How do I install groove-utilities-stats in Claude Code?
Run npx skills add andreadellacorte/groove --skill groove-utilities-stats -a claude-code. The skill lands at ~/.claude/skills/groove-utilities-stats/SKILL.md.
What does groove-utilities-stats do?
It reports lessons captured versus graduated, how consistently daily logs and rollups are kept, and how many tasks and commits moved in the period. Output goes to the conversation only.
Why are task counts missing from the report?
Counts appear only when tasks.storage in .groove/index.md names a backend such as beans, linear or github. With none the dashboard shows commits alone.
Source text
SKILL.md · MIT
---
name: groove-utilities-stats
description: "Quantified compound-loop dashboard: lessons captured vs graduated, adherence streaks, and task velocity. Use to see whether the workflow is actually compounding."
license: MIT
allowed-tools: Read Glob Grep Bash(git:*) Bash(beans:*) Bash(gh:*) Bash(linear:*) Bash(jq:*) Bash(python3:*) Bash(chmod:*) AskUserQuestion
metadata:
author: andreadellacorte
bash: true
---
<!-- groove:managed — do not edit; changes will be overwritten by groove update -->
# groove-utilities-stats
Use $ARGUMENTS to specify the period: `week` (default), `month`, or `all`.
## Bash fast-path
This skill has a bash script at `scripts/stats.sh`. Run it directly for faster, model-free execution:
```bash
bash .agents/skills/groove-utilities-stats/scripts/stats.sh week
```
(Pass `week`, `month`, or `all` to match $ARGUMENTS.) If the script exits 0, report its stdout and stop — do not continue with the steps below. If it exits non-zero, continue with the markdown steps (exit 3 means it printed the local metrics but the `linear`/`github` task counts still need to be computed below; exit 1 means no `.groove/`).
## Outcome
A read-only dashboard is printed to the conversation quantifying compound-loop health for the period: the **compound funnel** (lessons captured vs graduated), **adherence** (daily-log coverage, streak, rollup freshness, ratings), and **velocity** (task open/closed, commits). No files are written. This answers "is the workflow compounding / am I keeping the rhythm" — for qualitative patterns, it points to `/groove-utilities-memory-retrospective`.
## Acceptance Criteria
- Compound funnel shown: lessons captured (total + in period), graduated (total + in period), graduation rate
- Adherence shown: daily-log coverage over the last `memory.review_days` business days, current streak, weekly/monthly rollup freshness, ratings count + average
- Velocity shown: task open/closed counts (if a task backend is configured) and commit count in the period
- Output goes to the conversation only; nothing is written
- Absent data sources are skipped with sensible "n/a"/"NA" placeholders rather than errors
## Steps
(Model fallback — the bash fast-path above does this deterministically. Reproduce the same output format.)
1. Memory path is always `.groove/memory/`. If `.groove/` is absent, stop with an install hint.
2. Read `.groove/index.md` frontmatter: `tasks.storage` (backend; legacy key `tasks.backend`) and `memory.review_days` (default 5).
3. Determine date range from $ARGUMENTS: `week` = last 7 days, `month` = last 30 days, `all` = everything. Call the start `<since>`.
4. **Compound funnel**:
- Captured: count `## YYYY-MM-DD` dated entries across `.groove/memory/learned/*.md` (exclude `signals.md`) — total, and those with date ≥ `<since>`.
- Graduated: count `[graduated YYYY-MM-DD]` markers across the same files — total, and those ≥ `<since>`.
- Graduation rate = graduated_total / captured_total as a percentage (`n/a` if none captured).
5. **Adherence**:
- Daily-log coverage: of the last `review_days` business days (Mon–Fri, ending today), how many have `.groove/memory/daily/YYYY-MM-DD.md`.
- Streak: consecutive most-recent business days (back from today) that have a daily log; stop at the first gap.
- Rollup freshness: `weekly` is `fresh` if a `.groove/memory/weekly/<ISO-year>-W<ISO-week>.md` exists for the current or previous week, else `stale`; `monthly` likewise for `.groove/memory/monthly/YYYY-MM.md`.
- Ratings: parse `.groove/memory/learned/signals.md` table rows (`| YYYY-MM-DD | N/5 | note |`) with date ≥ `<since>`; report count and average (and a `▁▂▄▆█` sparkline if ≥ 3).
6. **Velocity**:
- Commits: `git log --oneline --since=<since>` count.
- Tasks, by `tasks.storage`:
- `beans`: open = `beans list --json --no-status completed --no-status scrapped | jq length`; closed = `--status completed` + `--status scrapped`; best-effort open mistakes/promises = count of non-terminal children under the "Mistakes"/"Promises" epics.
- `linear`: via the linear CLI/MCP — assigned issues grouped into open vs done.
- `github`: `gh issue list --assignee @me --state all` grouped into open vs closed.
- `none`: omit task counts (show commits only).
7. Print the dashboard in the **Output format** below.
## Output format
```
## Groove Stats — <period> (<start> to <today>)
### 🔁 Compound funnel
Lessons captured: <N> (<n> this <period>) Graduated: <G> (<g> this <period>)
Graduation rate: <pct>
### 📿 Adherence
Daily logs: <x>/<review_days> business days Streak: <s> days
Rollups: weekly <fresh|stale>, monthly <fresh|stale>
Ratings: <c> in period | avg <a>/5 <sparkline>
### 🚀 Velocity
Tasks: <open> open / <closed> closed Commits: <k> in period
Open: <m> mistakes, <p> promises
_For qualitative patterns and learnings, run `/groove-utilities-memory-retrospective`._
```
## Constraints
- Read only — never write to any file (no memory, no AGENTS.md, no config).
- Skip any absent data source with a one-line `n/a`/`NA` rather than erroring; parse tables and dates defensively.
- Paths are hardcoded under `.groove/memory/` (not configurable), consistent with the rest of groove.
- This skill complements `/groove-utilities-memory-retrospective` (qualitative reflection) — do not duplicate its narrative; cross-link it.
© andreadellacorte · MIT · andreadellacorte/groove