Claude Skill

os-say-simple

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 c

LLM Mart · 0 points · 2 views 0 listing impressions 0 install-command copies
Virus-scanned Reviewed automatically before listing.

Full trust report

Download kharmanskyi-open-steps-skills_os-say-simple-94b1cba.zip · 3 KB
Part of kharmanskyi/open-steps — 7 skills

Install

skills CLI npx skills add https://github.com/kharmanskyi/open-steps/tree/main/skills/os-say-simple
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install kharmanskyi-open-steps@llmmart
Git git clone https://github.com/kharmanskyi/open-steps.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole kharmanskyi/open-steps collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

os-say-simple

A text exists and the user cannot use it. This skill rewrites it into words they can act on - the pack's rescue skill: the others fire at moments of work, this one whenever any text, including your own, needs saying again properly. It changes only the words. Never the facts.

Language

Write in the language the user speaks in this session. Detect it from the conversation. Keep code, file names and identifiers in English.

When to use

  • The user asks for simpler or shorter, in any phrasing: "say it simply", "in plain words", "what does this mean", "I don't understand", "too long", "wait, what?", "bro".
  • The user pastes text and asks what it says.
  • The user gives the skill a number - they want exactly that many points.

Not this skill:

  • "Is this true?" is verification - os-check-work; a restatement stays exactly as true as its source.
  • "What do I do?" right after an answer that asked the user to act is a walkthrough - os-step-by-step. Asked about the road ahead, it is os-whats-next.

What text to restate

  1. Whatever the user points at: pasted text, a named file, a report in the reports folder. Read it if it is a file.
  2. Nothing pointed at: your own last substantial answer. A bare "wait, what?" always means the thing you just said.

The controlled-language rules

Drawn from the ideas behind ASD-STE100, the Simplified Technical English standard from aerospace manuals. This pack follows the ideas; it does not claim conformance to the standard.

  • One sentence, one idea. In instructions: one action per sentence.
  • Active voice. Name who does what - "the check refused the deploy", never "the deploy was refused".
  • Short sentences. A sentence that needs a breath in the middle is two.
  • One word, one meaning - and one meaning, one word. Never rotate synonyms for the same thing; the reader will assume the difference is meaningful.
  • Present tense unless the time itself is the fact.
  • Lists over dense paragraphs - but a list is not permission to drop the connections between items.

These rules shape every restatement - and when a term has to stay (rule 6), the sentences around it obey them with the term kept exact.

With a number

3 means exactly three points. Most important first, one sentence each.

  • Fewer real points than asked for: give the real ones and say the source holds no more. Never pad.
  • A warning or a risk is always one of the points. It outranks good news.

Hard rules

  1. Add nothing, drop no bad news. Every warning, risk, number and caveat in the source survives the rewrite. A summary that loses the ⚠ line is a lie by omission. Security, data loss and anything hard to undo get spelled out in full - the pack-wide exception applies here too.
  2. Numbers stay exact. No rounding money, counts or dates. A ticket or pull request number that names an action stays.
  3. As true as the original, no truer. The source's claims stay claims: "the report says the tests pass", not "the tests pass" - unless this session verified it. Offer os-check-work when the user needs true.
  4. Unclear stays unclear. Where the source is ambiguous, say "the original does not say" - never resolve ambiguity by guessing.
  5. Lead with the point. No preamble, no "here is a simpler version" - the first line is already the answer.
  6. A term with no plain equivalent gets described by what the user would see or do, with the term once in brackets - same move as the whole pack.
  7. Code and commands are exact strings. Never paraphrase what someone would type or run; explain around it.
  8. Aim for half the length or less - except where rule 1 or an exact-N request forbids it.

Known gotchas

  • "I don't understand" about your own answer is feedback, not an attack. Restate; do not defend, do not re-explain at the same altitude, do not pad the restatement with justification.
  • Restating is not editing. If while restating you notice your original was wrong, that is a correction - say so plainly. Never fix it silently inside the "simpler" version.
  • Mid-task "wait, what?" - restate, then continue the task. Do not lose the thread the question interrupted.
  • A text that is already plain gets one line saying so and nothing else. Manufacturing a different version to look useful is padding.
  • No number given, long source - the one-screen rule from the rest of the pack applies: about ten lines, fifteen at the ceiling.
Files (open-steps)
  • references
    • 01-wall-of-text.md 1.7 KB
      # Example 1 - a status update nobody outside engineering can use
      
      Source: a real end-of-day message, anonymised. Two calls against the same
      text.
      
      ## The text
      
      > Rebased onto main and force-pushed after the upstream squash-merge
      > invalidated the branch ancestry. CI green except the flaky e2e lane (known,
      > tracked). Bumped starlette to close the transitive CVE; lockfile
      > regenerated. The reconcile worker now short-circuits on idempotency-key
      > collision instead of raising, so Monday's retry storm can't recur. Left the
      > feature flag off pending the grants ceremony.
      
      ## `os-say-simple`
      
      > Monday's failure can't happen again: when the same request is submitted
      > twice, the system now keeps one copy instead of crashing. A security hole
      > in one of our libraries is closed. All automatic checks pass except one
      > that is known to be unreliable. None of this is live yet - it waits behind
      > a switch, until the database-permissions step that only you can run.
      
      ## `os-say-simple 3`
      
      > 1. Nothing is live yet - it all waits on the permissions step that is yours.
      > 2. Monday's double-submission failure is fixed for good.
      > 3. A known security hole in one library is closed.
      
      ## What the rewrites did
      
      | Move | Where |
      |---|---|
      | Branch mechanics dropped entirely | rebase and force-push change nothing for this reader |
      | The caveat survived every version | "not live yet" is bad news, and in the 3-point version it ranks first |
      | The not-quite-green status survived | "except one known unreliable check" - shortening may not turn amber into green |
      | Terms became what the reader would see | "retry storm" → "crashing on a doubled submission" |
      | The claims stayed the source's | nothing here was verified by the rewrite - as true as the message, no truer (rule 3) |
      
  • SKILL.md 5.1 KB
    ---
    name: os-say-simple
    description: >-
      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.
    allowed-tools:
      - "Read(~/.claude/open-steps/**)"
    ---
    
    # os-say-simple
    
    A text exists and the user cannot use it. This skill rewrites it into words
    they can act on - the pack's rescue skill: the others fire at moments of work,
    this one whenever any text, including your own, needs saying again properly.
    It changes only the words. Never the facts.
    
    ## Language
    
    Write in the language the user speaks in this session. Detect it from the
    conversation. Keep code, file names and identifiers in English.
    
    ## When to use
    
    - The user asks for simpler or shorter, in any phrasing: "say it simply",
      "in plain words", "what does this mean", "I don't understand", "too long",
      "wait, what?", "bro".
    - The user pastes text and asks what it says.
    - The user gives the skill a number - they want exactly that many points.
    
    Not this skill:
    
    - "Is this true?" is verification - `os-check-work`; a restatement stays
      exactly as true as its source.
    - "What do I do?" right after an answer that asked the user to act is a
      walkthrough - `os-step-by-step`. Asked about the road ahead, it is
      `os-whats-next`.
    
    ## What text to restate
    
    1. Whatever the user points at: pasted text, a named file, a report in the
       reports folder. Read it if it is a file.
    2. Nothing pointed at: your own last substantial answer. A bare "wait, what?"
       always means the thing you just said.
    
    ## The controlled-language rules
    
    Drawn from the ideas behind ASD-STE100, the Simplified Technical English
    standard from aerospace manuals. This pack follows the ideas; it does not
    claim conformance to the standard.
    
    - One sentence, one idea. In instructions: one action per sentence.
    - Active voice. Name who does what - "the check refused the deploy", never
      "the deploy was refused".
    - Short sentences. A sentence that needs a breath in the middle is two.
    - One word, one meaning - and one meaning, one word. Never rotate synonyms
      for the same thing; the reader will assume the difference is meaningful.
    - Present tense unless the time itself is the fact.
    - Lists over dense paragraphs - but a list is not permission to drop the
      connections between items.
    
    These rules shape every restatement - and when a term has to stay (rule 6),
    the sentences around it obey them with the term kept exact.
    
    ## With a number
    
    `3` means exactly three points. Most important first, one sentence each.
    
    - Fewer real points than asked for: give the real ones and say the source
      holds no more. Never pad.
    - A warning or a risk is always one of the points. It outranks good news.
    
    ## Hard rules
    
    1. **Add nothing, drop no bad news.** Every warning, risk, number and caveat
       in the source survives the rewrite. A summary that loses the ⚠ line is a
       lie by omission. Security, data loss and anything hard to undo get spelled
       out in full - the pack-wide exception applies here too.
    2. **Numbers stay exact.** No rounding money, counts or dates. A ticket or
       pull request number that names an action stays.
    3. **As true as the original, no truer.** The source's claims stay claims:
       "the report says the tests pass", not "the tests pass" - unless this
       session verified it. Offer `os-check-work` when the user needs true.
    4. **Unclear stays unclear.** Where the source is ambiguous, say "the
       original does not say" - never resolve ambiguity by guessing.
    5. **Lead with the point.** No preamble, no "here is a simpler version" -
       the first line is already the answer.
    6. **A term with no plain equivalent** gets described by what the user would
       see or do, with the term once in brackets - same move as the whole pack.
    7. **Code and commands are exact strings.** Never paraphrase what someone
       would type or run; explain around it.
    8. **Aim for half the length or less** - except where rule 1 or an exact-N
       request forbids it.
    
    ## Known gotchas
    
    - **"I don't understand" about your own answer is feedback, not an attack.**
      Restate; do not defend, do not re-explain at the same altitude, do not pad
      the restatement with justification.
    - **Restating is not editing.** If while restating you notice your original
      was wrong, that is a correction - say so plainly. Never fix it silently
      inside the "simpler" version.
    - **Mid-task "wait, what?"** - restate, then continue the task. Do not lose
      the thread the question interrupted.
    - **A text that is already plain** gets one line saying so and nothing else.
      Manufacturing a different version to look useful is padding.
    - **No number given, long source** - the one-screen rule from the rest of the
      pack applies: about ten lines, fifteen at the ceiling.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.

Related