open-loops
Skill2 starsChanged 24 days agoMITView on GitHub
Find email threads where someone is waiting on you (24h+) or you are waiting on someone (72h+). Read-only, deterministic rules, zero LLM judging.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- haorantang97/LabKittheir own version, first seen Nov 1, 2026
- trac3r00/agent-skillstheir own version, first seen Aug 26, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- haorantang97/LabKit★ 9same name
- trac3r00/agent-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository57 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- advisorSkill
- agent-dispatchSkill
- arenaSkill
- audit-pypi-model-packageSkill
- branching-workflowSkill
- code-reviewSkill
- criticSkill
- cron-child-oauth-auditSkill
- daily-forestSkill
- deep-researchSkill
- diagnose-cron-timeoutSkill
- dreamingSkill
- eod-summarySkill
- filingSkill
- forest-synthesisSkill
- goalsSkill
- graph-hygieneSkill
- ideaSkill
- inline-tagsSkill
- intel-analyzerSkill
- longformSkill
- loop-experimentSkill
- loop-librarySkill
- meta-memory-reviewSkill
- office-hoursSkill
- plan-ceo-reviewSkill
- plan-design-reviewSkill
- prdSkill
- progressive-disclosureSkill
- reality-feedback-createSkill
- reality-review-weeklySkill
- recallSkill
- reconcile-extractionSkill
- redactSkill
- repo-auditSkill
- research-digestSkill
- review-gateSkill
- run-long-job-detachedSkill
- saveSkill
- scopeSkill
- self-auditSkill
- shadow-test-local-classifierSkill
- signal-scorerSkill
- skill-auditSkill
- skill-chainingSkill
- spawn-agent-in-tenantSkill
- spawn-tenantSkill
- split-log-rows-by-session-kindSkill
- subagent-delegationSkill
- telegram-dump-routerSkill
- telegram-reply-toolSkill
- verify-stackSkill
- wave-planSkill
- weekly-retroSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 13
Search with newer_than:14d (in:inbox OR in:sent) -category:promotions -category:social -category:updates -category:forums, max 100 results.
- Distinctive: line 14
Keep the FIRST message id per threadId (that is the thread's latest message).
- Very specific: line 17
Write {"me": ["you@example.com"], "messages": [...metadata objects...]} to a scratch file.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Open loops
Pattern from gbrain docs/guides/open-loops.md. Run it inside a morning briefing, or on demand. Read-only: it never drafts, sends or labels anything.
Fetch (in-session, through your mail MCP's search and get tools)
- Search with
newer_than:14d (in:inbox OR in:sent) -category:promotions -category:social -category:updates -category:forums, max 100 results. Results come newest first. - Keep the FIRST message id per threadId (that is the thread's latest message). Cap at 30 threads.
- Get each kept id as metadata (headers + snippet), in one parallel batch.
- For any thread whose latest message is yours, also fetch the previous message id from step 1, so a nudge does not reset the clock.
- Write
{"me": ["you@example.com"], "messages": [...metadata objects...]}to a scratch file. Or setOPERATOR_EMAILSand leavemeout.
Judge
python3 -I scripts/memory/open_loops.py <input.json> (add --json for machine output).
- Exit 2 with
OPEN_LOOPS_EMPTY_INPUTmeans the fetch returned nothing. That is a FAIL, never "no open loops". Re-fetch. - Self-check:
python3 -I scripts/memory/open_loops.py --selftest(14 cases).
Rules (all in the script)
- Waiting on you: last real message is theirs, you are in To:, it asks or requests something, 24h+ old.
- You waiting: last real message is yours, it has a "?" outside any URL, 72h+ old.
- Acks ("thanks", "got it", "noted") never close a loop. Same-side nudges keep the original clock.
- Ignored: automated senders, calendar invites, promo/social/update/forum categories, threads older than 14 days, and senders in
memory/state/open-loops-ignore.json({"senders": ["domain.example"]}). - Writing in another language? Add its ack and request phrases to
ACKandREQUESTin the script.
What a result does NOT prove
It is judged on the ~200-char snippet, so a request deep in a long mail can be missed. Only the 30 most recent threads are checked. Follow-ups that moved to a chat app are invisible: a thread that moved there can show as open.
Report
List "waiting on you" first, oldest first, max 5 lines, then "you're waiting on them". When you dismiss a sender as never-a-loop, add the domain to the ignore file (safe to do without asking: it only narrows output).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

