本文へ移動
cccskills
無料GitHub で公開

os-step-by-step

ALWAYS invoke this skill when you need the user to act - run a command, paste a secret, click, approve - and whenever they ask how to do something or say they do not know what to do: "step by step", "walk me through it", "what do I do", "what should I do", "I don't understand what to do", "explain what I need to do", in any language. Picking which task comes next is os-whats-next; this skill is for doing the thing in front of you. First earn the ask: try it yourself, find another route, shrink it to the part only they can do. Then one action per step, commands labelled by what they touch, no jargon. Commands are single lines that prompt for any value - typing hidden for secrets - and confirm in plain words. Afterwards verify their step.

インストール方法を見る

含まれるファイル(2)

  • SKILL.md7.4 KB
  • references/01-paste-a-secret.md4.0 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

os-step-by-step

The user is not stuck because the task is hard - they cannot tell what they are being asked to do. This skill turns "I need something from you" into an instruction a person who does not read code can follow without a follow-up question. os-done-or-not reports that something is needed; this one says exactly how.

Language

Write in the language the user speaks in this session, detected from the conversation. Commands, file names and identifiers stay English.

When to use

Triggers live in the description above - any moment you need the user's hands, or they ask what to do.

Step 0 - earn the right to ask

Every ask costs the user a context switch. Prove it is necessary; stop at the first item that clears the block:

  1. Try it. A real 403 is a finding; "I probably lack permission" is a guess.
  2. Find another route. Another tool, a value already on the host, a file you can read.
  3. Shrink the ask. Obtain what you can yourself; hand over only the irreducible part.
  4. Check you are not asking twice - search the session and the reports folder first.

Three walls where asking IS the correct move, never to be worked around: pulling a secret into your own context, loosening a guard that is there on purpose, doing what the user said only they do.

The shape

Always this order - the ask first, never after the diagnosis.

**What I need from you: <one sentence, plain words>.**

Why you and not me: <one or two sentences. A real reason, in human terms.>

**Step 1. <action in three to six words>**
<What to do. One action only.>

**Step 2. <action>**
<...>

**How you'll know it worked.**
<What the user will see. What to do if they see something else.>

**What happens next, on my side.**
<One line: what you do once they are done, and what you will say.>

After they act - verify, do not trust

"Done" is a claim. Run the one quickest check that would fail if the step had not worked - the file exists with the right owner, the service answers, the value works once. Never print a secret to confirm it: confirm its effect. If the check fails, give only the corrected step - never the whole list again. Verified versus assumed is exactly what os-done-or-not needs for "yes" versus "not checked".

Secrets and dangerous steps

A secret never goes through the chat - it would stay in the history and the logs. The command below puts it where it belongs directly; say who deletes it and when. A hard-to-undo step - live users, money, deletion - gets its own warning line before the command.

Writing a secret to a remote host - one line: it prompts, hides the typing, refuses a truncated paste, confirms by size:

printf 'Paste the connection string, then press Enter: '; IFS= read -rs V; echo; if [ ${#V} -lt 20 ]; then echo "Only ${#V} characters - that looks truncated, nothing was saved."; else printf '%s' "$V" | ssh root@HOST 'umask 077 && cat > /path/to/secret' && ssh root@HOST 'echo "Saved, $(wc -c < /path/to/secret) bytes"'; fi; unset V

Typed by hand rather than pasted - ask twice; a typo in a hidden field is otherwise undetectable:

printf 'Enter the token: '; IFS= read -rs A; echo; printf 'Enter it again: '; IFS= read -rs B; echo; if [ "$A" != "$B" ]; then echo "The two entries differ - nothing saved, run it again."; else printf '%s' "$A" | (umask 077; cat > /path/to/secret) && echo "Saved, $(wc -c < /path/to/secret) bytes"; fi; unset A B

These templates are for bash and zsh and have not been run on Windows. There, ask the user to open a separate Git Bash window just for this command, and to leave the agent running where it is. Never write an unchecked PowerShell one.

Two properties the templates cannot keep for you, both measured:

  • printf …; IFS= read -rs VAR - never read -rsp. In zsh -p means "read from a coprocess": the variable comes back empty with no error and the secret file is written blank. macOS defaults to zsh.
  • The value never appears in the command itself - only piped from the variable; anything in the arguments lands in shell history and the process list.

The rest the templates already embody - one single line, umask 077 before writing, refuse short input, confirm by byte count never by content, unset at the end. Adapt the prompt, the threshold and the path; keep every property.

The same pattern serves any value the user must supply by hand - a public key, a domain, an address, an id. Prompt for it the same way; keep the typing visible when the value is not secret (drop -s), skip the length gate when short is valid - and always end with a plain-words confirmation of what just happened, so pressing Enter never feels like dropping a coin into a well.

Choices, not instructions

A decision gets no steps. Use your tool's question picker where it has one (in Claude Code, AskUserQuestion), with the pack's contract: plain question, why it matters, what changes later, easy to undo, two to four options, the recommended one first and marked. Where there is no picker, write the same question and options as plain text, the recommended one first and marked.

Hard rules

  1. The ask goes first. What you tried and what failed is your problem - one line at the end, or nothing.
  2. One action per step. Two commands is two steps.
  3. Label every command with what it touches - the test server, the live server, their own machine. Look-alike steps on different targets: say what happens if they are swapped.
  4. A command is self-contained. One line the user pastes and runs; if it needs a value, it asks for it. Never make the user feed input by redirection, a heredoc or Ctrl-D.
  5. No jargon inside a step. Avoid terms instead of explaining them: write what the person sees and clicks. One unavoidable term may stay - without a lecture.
  6. Always give a way to check. A step the user cannot verify is a step they will redo out of doubt.
  7. Never mix your work with theirs. Two lists with plain headings; a buried "this one's on you" is not an instruction.
  8. Never compress a multi-step sequence - here brevity causes misreads. Blocks of two to three sentences; longer gets skimmed, and a skimmed step is a missed step.

Known gotchas

  • A heading is not a summary: "one command per host" above two commands reads as one command. Count out loud.
  • A dropped step is "skip step 3" - never a silent renumber.
  • If you can verify it yourself, verify it yourself; do not ask for confirmation you do not need.
  • "Say done" needs a subject - the user may have three of your requests open.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

ALWAYS invoke this skill before asking the user any technical question or offering options, and whenever they ask to be asked in plain words - "ask simple", "ask me simply", "ask me in plain words" - in any language. ALWAYS invoke it too when the user asks for your view on a technical choice: whether it is worth doing, more than the problem needs, or replaceable by something simpler, and which option you would take. Invoke it even when the code makes the answer look obvious: the checks are what make the answer more than a guess. Rewrites the question in plain words and always ends with one marked recommendation. A structural choice first passes six checks, shown as a table: effort now, simpler substitute, extra work later, lock-in, over-engineering, easy to undo. Doing nothing is always weighed.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

ALWAYS invoke this skill when the user asks where the project as a whole stands - "where are we", "what's the big picture", "what have we built", "what is in this project", "map the project", "what does this thing even do", "what is stale", "what can we delete", "update the roadmap" - in any language, and whenever a session report has just been written. It keeps one file, BIG-PICTURE.md: what the product is, every feature with how far it got, which parts nobody has touched, and what is queued next. Age and wiring are measured from git; how far a feature got comes from the reports or says "not checked". Where a tracker is connected it offers to open the queue as tickets, after a yes and never before. Never invents work, never deletes code.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

ALWAYS invoke this skill when the user asks about work done outside this session - "check work", "check the others", "check other sessions" - or to accept one: "the session is done, check it", "can we merge it" - in any language. You are the receiving party: treat the report as a claim, verify each part against machine state, name every gap between claimed and true. Verified-ready work merges in the same pass. End with what is left, one next step, and any reply another session needs, ready to paste.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

ALWAYS invoke this skill when work wraps up or the user asks how it went - "done or not", "are we done", "what happened", "report", "what's the status of this ticket" - in any language, and when a Stop hook asks for a session report. Produces a ten-line plain- language report: a lead, a checkmark table, and a verdict - fully done, anything needed from you, new debt, safe to close. Every "yes" names its proof; unverified says "not checked". Saves the report so the next session starts from it instead of re-exploring the repo.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

ALWAYS invoke this skill when the user asks for simpler or shorter about something said or written - "say it simply", "what does this mean", "I don't understand your answer", "too long", "wait, what?", "bro" - in any language, about any text: your own answer, a report, a review comment, an error. "I don't understand what to DO" is os-step-by-step; this skill restates text. It restates for a reader who does not read code: leads with the point, keeps every number, warning and caveat - no facts added, no bad news dropped. A number returns exactly that many points, most important first.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

ALWAYS invoke this skill before anything hard to undo gets agreed to - a contract, a purchase, a migration, a launch, a price change, a reorganisation - and whenever the user asks "what could go wrong", "what are we missing", "poke holes in this", or for a premortem or a red team, in any language. Assumes the decision already failed and works backwards to find out why, in a fresh agent that had no hand in making it. Sweeps nine areas and shows what each produced, including the empty ones. Ends on one verdict: go ahead, go but fix these first, try it small first, think again, do not do this.

日本語の概要は準備中です。原文の説明を表示しています。

kharmanskyi/open-steps1,2822026年10月6日 更新

kharmanskyi のスキルをすべて見る

このスキルの問題を報告する