quick-do
Skill11 starsChanged 12 days agoMITView on GitHub
Execute small, well-understood changes directly in the current session with minimal ceremony and ponytail taste. Use when explicitly requested, or automatically for localized fixes, mechanical edits, config tweaks, and short scripts with clear behavior and a known verification path. No plan documents or grilling; do not add tests by default. Prefer steady-do when design, compatibility, or acceptance decisions remain; prefer plan-brief when planning and execution need a separate-session handoff.
In this repository39 other files it gives its agents.
- accel-initSkill
- accel-platform-installSkill
- accel-pytorch-code-portingSkill
- accel-pytorch-pythonSkill
- accel-skill-templateSkill
- model-datasetsSkill
- platform-environment-skill-auditSkill
- agent-reach-installSkill
- code-reviewSkill
- consolidatorSkill
- diagnosing-bugsSkill
- distill-project-kbSkill
- domain-modelingSkill
- editable-vector-slidesSkill
- goal-firstSkill
- grillingSkill
- image-genSkill
- load-memSkill
- migrate-memSkill
- personal-ui-tasteSkill
- plan-briefSkill
- project-profileSkill
- readable-docsSkill
- save-memSkill
- steady-doSkill
- tddSkill
- teachSkill
- writing-skillSkill
- kb-8bit-quantizationSkill
- kb-accel-sdkSkill
- nju-pptSkill
- exp-batchSkill
- exp-campaignSkill
- exp-discussSkill
- exp-probeSkill
- exp-workflowSkill
- kernel-workerSkill
- pytorch-perf-tuningSkill
- research-evidenceSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Read the request, applicable project rules, and affected files before editing.
- Distinctive: line 21
Prefer the obvious interpretation for ordinary implementation details.
- Distinctive: line 22
No plan documents, grilling sessions, or sub-agents.
- Distinctive: line 23
Keep the smallest working change: reuse existing code, then stdlib/native features, then installed dependencies.
- Distinctive: line 24
Before destructive or irreversible actions, show the impact and obtain explicit authorization.
- Distinctive: line 28
Default to the smallest real execution path plus a reread of the change.
- Distinctive: line 29
Do not add tests, TDD, or verification scripts by default.
≈ “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
Quick Do: understand → do → verify
Apply goal-first: keep the requested outcome and smallest sufficient check in view. An incidental discovery is work only if it affects that outcome; do not turn a bounded edit into exploration or cleanup. No extra planning artifact.
When this skill applies
- Small, well-understood changes with a known verification path: localized fixes, mechanical edits, config tweaks, short scripts.
- The user's explicit choice wins — never switch workflows silently; if the chosen one cannot fit the task, say why and ask. Stuck on design, compatibility, or acceptance decisions → stop and propose
steady-do(if unsure between quick and steady, use steady); a handoff to a separate session → proposeplan-brief. - These workflows govern requested development work, not every explanation or read-only question. Skill instructions are in English; replies and deliverables follow the user's language (Chinese by default).
Do the work
- Read the request, applicable project rules, and affected files before editing. Inspect relevant callers when changing shared behavior; avoid unrelated exploration. Reuse unchanged context, keep the exact edit unit available, and refresh it after a match failure instead of guessing another patch.
- Look at the live site first (seconds, no approval): env active, deps actually importable, files and callers as expected — one glance that prevents a whole round of avoidable errors. Skipping ceremony is the point here; skipping this look is not.
- Fix the root cause, not the reported symptom: grep the callers of what you touch and fix where all of them route through — a guard repeated in every caller is a bigger diff and leaves siblings broken.
- Prefer the obvious interpretation for ordinary implementation details. Ask at most one focused clarification in the normal quick path; if investigation reveals substantial ambiguity or risk, stop and propose
steady-dorather than guessing to satisfy the question limit. - No plan documents, grilling sessions, or sub-agents. Use a todo only when the work or host instructions warrant it; do not create a planning ceremony.
- Keep the smallest working change: reuse existing code, then stdlib/native features, then installed dependencies. No over-encapsulation, speculative abstractions, defensive boilerplate, redundant checks, or unrelated cleanup. Preserve necessary trust-boundary validation, security, accessibility, and data protection.
- Before destructive or irreversible actions, show the impact and obtain explicit authorization. A quick workflow does not permit overwriting secrets or changing unrelated environments.
Verify and finish
- Default to the smallest real execution path plus a reread of the change; check the requested behavior, not merely the absence of errors. For documentation-only changes, inspect content and applicable format checks.
- Do not add tests, TDD, or verification scripts by default. Follow explicit user requests and mandatory project checks; those are not waived by this workflow.
- Report only what actually ran. A static check, successful load, or injected input is not proof of end-to-end behavior. If execution is unavailable, state what remains unverified and the concrete user check.
- Blocked is a legitimate outcome: say so, name what is left and the exact check the user must run; never fake completion and never silently skip a step.
- Finish in one short line stating the change and verification status, including a blocker or required user check when present. No separate report or summary essay; required product docs and project memory still apply.
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.

