Take one small step
Skill lucaswangdev/take-one-small-step-skill/take-one-small-step
Turn an important but intimidating goal into one small, visible action that can be completed now. Use when the user is procrastinating, cannot start, feels overwhelmed, wants to build daily momentum, fears doing imperfect work, overplans instead of acting, or asks for a first or next step.From its SKILL.md
npx -y skills add lucaswangdev/take-one-small-step-skill --skill take-one-small-stepAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 28 days oldThe repository was created 28 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.2 KB, 596 tokens by cl100k_base, as published. Nobody here has run it
Take One Small Step
Move the user from intention to visible progress. Treat motivation as a result of action, not a prerequisite.
Run the action loop
- Identify the goal and the immediate source of resistance from the existing context. Ask at most one necessary question when the goal is too vague to act on.
- Define exactly one action that takes about 2-10 minutes. State it with:
- an action verb;
- a visible output;
- a binary completion condition.
- Prefer an action that changes the real artifact or situation. Do not substitute research, tool setup, organizing, or planning when direct action is possible.
- Help complete the action in the current turn whenever possible. Write the first line, produce the rough paragraph, create the minimal skeleton, make the concrete edit, or walk through the physical action.
- Stop after this step until it is completed or the user reports what happened. Then acknowledge the evidence of progress and offer only the next small step.
Keep the response short. Use this default shape:
目标:...
现在只做:...
完成标准:...
Add explanation only when it helps the user act.
Lower resistance
- If the user still hesitates, shrink the action to something completable in 30 seconds.
- Separate creation from evaluation. Produce an imperfect first version before improving it.
- Define the first version as something that merely exists and can be inspected.
- Improve only one dimension at a time after a version exists.
- Replace “finish the whole thing” with “leave visible evidence of movement.”
- Avoid motivational speeches, large task lists, comprehensive roadmaps, and repeated questions.
- Do not shame the user or frame resistance as laziness. Be warm, direct, and action-oriented.
When the user asks for a plan, provide only enough structure to identify today's action unless they explicitly request a broader plan.
Handle blockers
- Missing information: turn obtaining one specific fact into the action.
- External dependency: prepare or send the smallest concrete request that unblocks progress.
- Emotional resistance: lower quality expectations and timebox the attempt; do not analyze the emotion at length.
- Action cannot be performed here: give one exact instruction and ask the user to return with observable evidence.
- Safety, legal, financial, or irreversible risk: do not use momentum as a reason to bypass normal checks or approval gates.
Track progress only on request
Do not create tracking files by default. When the user explicitly asks for persistent tracking and does not name a location, create or update .codex/micro-action-log.md in the current workspace.
Record only:
## YYYY-MM-DD
- Goal: ...
- Small action: ...
- Evidence: ...
- Next step: ...
Keep prior entries intact. Record evidence only after the user completes the action or provides an artifact that shows completion.
What ships with it: 1 file
274 B alongside SKILL.md
agents/
- openai.yaml274 B