agentleFS
gleicongleicon/devskills/skills/ds-step-mode/SKILL.md

ds-step-mode

Skill20 starsChanged 46 days agoMITView on GitHub

Activate step mode for this session.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    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
  1. Distinctive:

    After completing a step, stop and report concisely: what you did, what changed (files/commands), and the next step you propose.

    line 13
  2. Distinctive:

    When you present options, list them as prose recommendations the user can accept, amend, or combine — never force a single selection.

    line 21
  3. Distinctive:

    Keep the plan honest: mark steps done as they complete.

    line 27

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

When active, execute work in small, user-gated steps instead of long autonomous runs. Do the smallest meaningful, reviewable step, then stop and hand control back so the user can approve, steer, change the plan, or checkpoint before the next step. The user drives; you take one step at a time. Invoked with a plan — /ds-step-mode current plan, a path like .project/roadmap.md, or pasted text — read it and work through it step by step with these breakpoints.

The loop

  • Work in the smallest meaningful, reviewable step — one logical change or action. Default to pausing more, not less; never silently chain multiple steps.
  • Propose before you act. State the single next step and wait — do not make the change until the user approves. This gives them a veto before anything happens.
  • After completing a step, stop and report concisely: what you did, what changed (files/commands), and the next step you propose. A few lines, not a wall of text.
  • Then yield: end your turn and wait. Do not start the next step until the user responds.
  • The user may approve ("go" / "next"), amend the next step, change the plan, ask for a checkpoint, ask a question, or redirect entirely — honor whatever they say, including "do the next three steps" (then pause again).
  • Granularity is tunable live: "bigger steps" / "smaller steps" adjusts how much you do before stopping. Start at one logical unit.

Handing control back (always steerable)

  • At any decision or handback point, return control in prose — end with a short, open question the user can answer freely. Never hand back via the multiple-choice picker: the user must always be able to accept an option and add instructions, or combine and redirect.
  • When you present options, list them as prose recommendations the user can accept, amend, or combine — never force a single selection.
  • The multiple-choice picker is acceptable only for a trivial either/or disambiguation (e.g. "did you mean file A or file B?") — never for "what should I do next" or steering a step.

Driving a plan

  • Restate the next unstarted step from the plan, then run the loop above on it.
  • Keep the plan honest: mark steps done as they complete; when the user changes course, update the plan to match. Don't rewrite the whole plan unprompted.
  • Offer a checkpoint (/ds-project-checkpoint) at natural milestones or on request, so the session stays safe to pause or /clear.

Confirm activation with "Step mode active." Activating a mode only turns on this posture; it is not approval to begin work — continue with whatever the user already asked for, or wait for their next instruction.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.