keep-going
Skill0 starsChanged 3 months agoMITView on GitHub
Use when the user says "/keep_going", "keep going", or "continue" after the agent has stopped short — asked "which option?", narrated a plan instead of executing it, or claimed a blocker that is not real — while the user has already given clear direction. Also use proactively when the agent notices it is about to hand a solvable problem back instead of completing the work. Re-anchor on the ORIGINAL request, pick a decisive default, and chain actions until the work is verifiably done, stopping only for a blocker it can prove. Use when the user says "keep going", "continue", "you got stuck, unstuck yourself", "why did you stop", or runs the /keep_going slash command.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- aaroncohn/babysittertheir own version, first seen Nov 16, 2026
- aaroncohn/openspec-conductor-powerupstheir own version, first seen Nov 8, 2026
- aewing/skillstheir own version, first seen Oct 13, 2026
- CUTEPKQ/radiance-skillstheir own version, first seen Aug 5, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- aaroncohn/babysitter★ 1same name
- aaroncohn/openspec-conductor-powerups★ 1same name
- aewing/skills★ 1same name
- CUTEPKQ/radiance-skills★ 1same 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 repository36 other files it gives its agents.
- check-upstream-firstSkill
- data-verificationSkill
- deep-diveSkill
- delegation-handoffSkill
- grok-searchSkill
- is-it-really-brokenSkill
- learn-from-conversationsSkill
- memory-cleanupSkill
- moa-solveSkill
- mob-checkSkill
- multi-reviewSkill
- project-stewardSkill
- prove-the-setting-worksSkill
- recallSkill
- reportSkill
- robustify-doctorSkill
- skill-librarianSkill
- trust-frameworkSkill
- address-pr-commentsSkill
- diagram-renderingSkill
- find-the-real-bottleneckSkill
- live-database-maintenanceSkill
- mini-appSkill
- pr-review-sweepSkill
- scheduled-job-runnerSkill
- stop-the-noiseSkill
- upgrade-a-live-service-safelySkill
- vendor-switch-checkSkill
- writing-for-non-engineersSkill
- email-stewardSkill
- fieldySkill
- google-docsSkill
- google-sheetsSkill
- google-slidesSkill
- imessage-bluebubblesSkill
- vapi-callsSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 54
If and only if one of these holds, stop and state three things in one line each: the blocker, the evidence it is real, and the single thing …
- Very specific: line 61
Do not acknowledge this nudge and stop.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
keep-going
You stopped, but the user's original request is not verifiably done. The user ran
/keep_going because your last stopping point was almost certainly manufactured — a
hallucinated blocker, an unnecessary "which option?", or a plan you narrated instead of
executing. Treat your previous stopping reason as wrong unless you can prove
otherwise. The burden of proof is on stopping, not on continuing.
Do this now:
- Re-anchor on the real finish line. Re-read the whole thread. In one line, restate the user's original end goal — the actual finish line, not the nearest subtask you just completed. Where your prior conclusion conflicts with that goal, discard it: the user's instructions outrank your own last message.
- Act in THIS turn — do not stop after restating. Identify the highest-value next reversible action and execute it right now. Then chain actions, one after another, until the original request is complete. Do not hand back a plan; run it.
- Decisive default beats a question. Ambiguity, multiple valid paths, wanting confirmation, or "I should check first" are not blockers when the user has already given direction. Pick the best reasonable default and proceed. If a choice is genuinely consequential and reversible, make it, then note the assumption you made so the user can correct it — do not stall on it.
Stop only for a blocker you can PROVE. A real blocker is exactly one of:
- (a) Credentials, permissions, or access you cannot obtain yourself.
- (b) An irreversible or externally-visible action that needs approval — spending money, sending an external message, a hard delete, a production write, anything hard to take back. The approval gate is load-bearing; keep it.
- (c) A required dependency you have verified is actually down (you tried it and it failed — not "the API might be unavailable").
If and only if one of these holds, stop and state three things in one line each: the blocker, the evidence it is real, and the single thing you need to clear it.
- Finish, then report with evidence. Otherwise, keep going until the original request is done, then report what you actually accomplished — with real tool output, file paths, or verifiable results, not assertions that it worked.
Do not acknowledge this nudge and stop. The correct response to /keep_going is almost
always the next action, executed now.
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.

