Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add andreadellacorte/groove --skill groove-work-compoundgit clone --depth 1 https://github.com/andreadellacorte/grooveWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/andreadellacorte/groove/groove-work-compound)<a href="https://agentmods.dev/skills/andreadellacorte/groove/groove-work-compound"><img src="https://agentmods.dev/badge/skills/andreadellacorte/groove/groove-work-compound/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/andreadellacorte/groove/groove-work-compound"><img src="https://agentmods.dev/badge/skills/andreadellacorte/groove/groove-work-compound.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- Socket pass
- Snyk pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00022 | $0.01347 |
| Opus 5 | $0.00011 | $0.00674 |
| Sonnet 5 | $0.00004 | $0.00269 |
| Haiku 4.5 | $0.00002 | $0.00135 |
Grade A, and why
groove-work-compound scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 67 lines — stays where its author put it; the contents beside it link to each section on GitHub.
groove-work-compound
Scope
"This conversation" means the full chat thread (all messages in the same chat where the user invoked compound). Lessons and compound actions come from work done or decided anywhere in that thread.
In scope:
- Work done earlier in the same chat: other skill runs (e.g. /ppp-weekly-report, /groove-work-plan), file or bean edits, decisions, review or planning discussion.
- Any bean or work item the user or the thread refers to as part of the workflow (e.g. the epic or task this compound stage belongs to). Use those to ground lessons and actions.
Not in scope unless the user asks: Daily memory, weekly summaries, or other docs (e.g. "compound on the whole day").
If no other work is visible in the context when compound runs (e.g. the turn only shows the compound request): ask which session or work to compound on (e.g. "I only see the compound request. Which session should I compound on — e.g. the PPP run above, or something else?") instead of concluding "no work discussed."
Outcome
Lessons, root causes, and fixes are documented. Relevant project files are updated (rules, templates, docs, task bodies). A checklist of compound actions with done/pending status is produced.
Acceptance Criteria
- Lessons are documented with root cause analysis (not just what went wrong, but why)
- Relevant project files are updated: rules files, templates, docs, task bodies
- Checklist is produced with each action marked done or pending
- If user frustration patterns were detected, they are specifically captured
- "No new lessons" is an explicit acceptable outcome — capture it rather than skipping
Constraints
- Read
tasks.backendfrom.groove/index.md; memory path is always.groove/memory/ - Pending capture: if
.groove/.cache/pending-capture.mdexists (staged by thesession-captureStop hook), read it first and use its commits/working-changes as ground-truth source material for the lessons and the git context — so a forgotten or compacted session is not lost. After the compound stage is recorded, delete.groove/.cache/pending-capture.md. - Before producing the compound actions checklist, check for open mistake incidents via
/groove-utilities-memory-mistakes --list:- If open incidents exist: process each using the log → fix → audit → resolve cycle from
/groove-utilities-memory-mistakesbefore moving on - Add to compound checklist: "mistakes: N incident(s) resolved → learned/" (done/pending)
- If no open incidents: skip silently
- If open incidents exist: process each using the log → fix → audit → resolve cycle from
- Output goes into existing project files — do not create new files unless necessary
- Always run even if it seems like "nothing to capture" — capture that explicitly
- Compound actions checklist must include why each action matters, not just what it is
- If user showed repeated fixes, confusion, or rework: capture the pattern and its trigger
- Create stage task in backend if
tasks.backend != nonevia/groove-utilities-task-createwith a descriptive title so multiple compounds per day are distinct:YYYY-MM-DD, Compound — <brief topic>(topic from the work just closed, e.g. release, feature, or session summary). Do not number stages in task titles. - After producing the compound actions checklist, identify any lesson that is about AI workflow, agent behaviour, tool usage, or engineering process — not specific to the codebase or product
- If any such lesson is found:
- Propose: "Workflow learning detected — add to .groove/memory/learned/.md?"
- Suggest a topic name based on the content (e.g.
patterns,tools,anti-patterns,hooks) - Wait for user to confirm, pick a different topic, or skip
- If confirmed: append the lesson as a bullet under a
## <YYYY-MM-DD>heading in the file; create the file with a# <Topic>heading if it does not exist; create the dated heading if absent - Add to compound checklist: "workflow learning → .groove/memory/learned/.md" (done/pending)
- If no workflow lessons, skip this step entirely — no prompt, no noise
- If a workflow learning was captured locally AND
groovebook:is set in.groove/index.md: add a one-line suggestion after the checklist: "Learning captured →.groove/memory/learned/<topic>.md. Publish to groovebook? Run/groove-groovebook-publish." Ifgroovebook:is absent: skip silently. - After a workflow learning is captured, offer graduation: "Is this lesson stable enough to apply to every future session? Run
/groove-utilities-memory-graduate <topic>to promote it to AGENTS.md permanently." — only suggest if the lesson was just confirmed, not every compound run. - After the workflow learning step, scan the conversation for deferred items — phrases like "we'll come back to", "do this later", "next time", "TODO", "skip for now", "won't fix today". If any are found: list them and ask "Capture any of these as promises? (numbers, or enter to skip)" — for each confirmed item, run
/groove-utilities-memory-promises <text>. If none found: skip silently. - After the workflow learning step, prompt for an optional session rating:
- Ask: "Rate this session (1–5): how well did the compound loop serve you? (enter to skip)"
- If the user provides a rating: append to
.groove/memory/learned/signals.mdunder a table; create the file with a header and table header row if it does not exist - Format:
| YYYY-MM-DD | <rating>/5 | <one-line context from the session topic> | - If the user skips: do not mention it again
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 67 lines · 22 tokens per session scan A fb1aea7b9730
groove-work-compound is a skill published in the GitHub repository andreadellacorte/groove (5 stars, last pushed 1mo ago), licensed MIT. It adds 22 tokens to every session and 1,347 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
orbit-notion
Open Orbit briefing skill — selected by the Orbit pipeline when Notion is the user's only connected connector, or when the user explicitly scopes their daily digest to Notion. Pulls the past 24 hours of document edits, comments, mentions, and database row changes from the user's authenticated Notion connection and…
Cortex
Operate Cortex, the LifeOS memory system — the typed Knowledge Archive (People, Companies, Ideas, Research with typed related: links) plus recall of prior work sessions, ISAs, and conversations. Search, add, harvest, develop, ingest, distill, graph-navigate, recall. USE WHEN cortex, knowledge, knowledge base, search…
pinchtab-mcp
Use this skill when a task requires browser automation through PinchTab's MCP server connected to a remote browser instance. Covers navigation, element interaction, data extraction, form filling, multi-step flows, and session management via MCP tools.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
peekaboo
Capture and automate macOS UI with the Peekaboo CLI.
mochi-remind
Handle due reminders — notify the user with natural language and mark them done.