agentleFS
berkcaparberkcapar/product-management-skills/skills/unslop/SKILL.md

unslop

Skill1 starsChanged 36 days agoMITView on GitHub

Cuts long, AI-generated PRDs and product docs down to the decisions, numbers and open questions. Use when asked to unslop, tighten, shorten or de-AI a PRD, spec, brief or product doc.

197 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use principle-never-block-on-the-human, greploop-apps and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • jaydubya818Skill

    principle-never-block-on-the-human

    Apply when safe reversible implementation details can proceed inside scope.

    Also in 45 projects that use this

  • jaydubya818Skill

    greploop-apps

    Run greploop through @greptile-apps when the normal trigger rejects a large PR. Use only with the companion greploop skill.

    Also in 5 projects that use this

  • ianclearySkill

    principle-explain-the-number

    Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.

    Also in 30 projects that use this

  • bnemaSkill

    setup-pstack

    Configure which Pi models pstack uses per role. Detects available provider/model specs and writes pstack's user configuration. Use for /setup-pstack, "configure pstack models", or changing pstack model choices.

    Also in 26 projects that use this

  • ianclearySkill

    poteto-help

    Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.

    Also in 15 projects that use this

  • +7 more from jaydubya818

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 17 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 192 edited versions we read keep these.

  • 17

    Distinctive

    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:

    Never invent: No new metrics, targets, dates, owners, segments or scope.

    line 14
  2. Distinctive:

    Decisions over prose: Every sentence that survives is a decision, a verifiable claim, or context the reader needs to make a decision.

    line 15
  3. Distinctive:

    Keep the source language: Write the output in the language of the input.

    line 17
  4. Distinctive:

    Your own output must pass the slop check: Run the pattern list below on the rewrite before returning it.

    line 18
  5. Distinctive:

    Read the whole doc and list, for yourself:

    line 23
  6. Distinctive:

    no success metric, or a metric with no baseline or target

    line 37
  7. Distinctive:

    no owner or timeline when the doc implies one

    line 41
  8. Distinctive:

    Use the template below. Reorder freely.

    line 44
  9. Distinctive:

    If a section has nothing in the source, write one line pointing to the open question (e.g.

    line 83
  10. Distinctive:

    After (source mentioned 60% day-one drop-off and a 7-screen flow):

    line 127
  11. Distinctive:

    After (source mentioned neither):

    line 130
  12. Distinctive:

    After the rewritten doc, add a short block:

    line 138
  13. Distinctive:

    If you had to make any interpretation call (e.g.

    line 151
  14. Distinctive:

    Don't add content, examples or metrics the source doesn't contain.

    line 159
  15. Distinctive:

    Don't soften open questions to be polite.

    line 160
  16. Distinctive:

    Don't preserve the source's section structure out of respect for it.

    line 161
  17. Distinctive:

    Don't end with a summary of what you did beyond the report.

    line 162

“Common” is measured: kept by at least 6 in 10 of the 192 edited versions of this file we could read. “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

Unslop

Turn a bloated product doc into the shortest version that keeps every decision, number and constraint. Make missing decisions visible instead of papering over them.

AI-generated PRDs have two problems. The obvious one is style: long sentences, filler words, rhetorical tricks. The deeper one is substance: the prose hides decisions nobody made. Fixing only the style gives you a shorter doc that is still empty. This skill fixes both.

Core rules

  1. Never invent. No new metrics, targets, dates, owners, segments or scope. If the source doesn't say it, it becomes an open question.
  2. Decisions over prose. Every sentence that survives is a decision, a verifiable claim, or context the reader needs to make a decision.
  3. Length follows decisions, not the original. A doc with three decisions does not need four pages. Never pad a section to make it look complete.
  4. Keep the source language. Write the output in the language of the input. Keep product and technical terms as the team uses them.
  5. Your own output must pass the slop check. Run the pattern list below on the rewrite before returning it.

Process

1. Extract before writing

Read the whole doc and list, for yourself:

  • decisions (what we will and won't do)
  • numbers (baselines, targets, sizes, dates)
  • constraints and dependencies
  • user segments
  • explicit non-goals

Everything not on this list is a candidate for deletion.

2. Sentence test

For each sentence in the source ask: is it a decision, a verifiable claim, or context the reader needs? If none, cut it. If it is a vague claim ("improve engagement", "better experience"), turn it into an open question: which metric, from what, to what?

3. Gap check

Compare what you extracted against the template sections. Each missing piece becomes an open question. Typical gaps:

  • no success metric, or a metric with no baseline or target
  • no non-goals
  • "users" with no segment
  • solution described, problem never evidenced
  • no owner or timeline when the doc implies one

4. Rewrite into the template

Use the template below. Reorder freely; do not keep the source's structure if the structure itself was padding.

5. Slop pass

Apply the pattern list to your rewrite. Then reread only the TL;DR: someone who reads nothing else must know what is being built, for whom, and what is still undecided.

6. Report

Append the short report described below.

Output template

# [Feature / project name]

**TL;DR:** max 3 sentences. What, for whom, why now.

Status: … · Owner: … · Target date: …

## Problem
What is broken and the evidence from the source. No evidence → open question.

## Who it's for
The specific segment.

## Non-goals
What we are explicitly not doing.

## Solution
What we are building. Scope, not aspirations.

## Success metrics
| Metric | Baseline | Target |
Missing values → open question, never a guess.

## Risks & dependencies

## Open questions
1. The question. Why it matters / what it blocks.

If a section has nothing in the source, write one line pointing to the open question (e.g. "Not specified, see Q2"). Do not delete the section silently; an empty section is information.

The status line carries what the source says about status, owner and date, word for word. A team instead of a person counts as "no named owner"; a missing date is "not set". Either one also becomes an open question.

Rough length target: under ~1 page (~500 words) for a typical PRD, usually 20–35% of the original. Docs with many real decisions can be longer. Docs with few decisions should be shorter.

Patterns to remove

Filler vocabulary

leverage, seamless, robust, empower, unlock, holistic, delve, streamline, cutting-edge, game-changer, best-in-class, world-class, synergy, ecosystem (when not literal), "in today's fast-paced world", "it's important to note", "it is crucial that", "at the end of the day".

Replace with the concrete thing, or delete.

Sentence level

  • One idea per sentence. Split or cut anything over ~25 words.
  • No restating. Drop section intros that announce what's coming and section endings that summarize what was just said.
  • No hedging. "Potentially", "may be worth considering", "in some cases", "could possibly". Either claim it or move it to open questions.
  • Actor + verb + object. "Improvements to the user experience are targeted" → "We're cutting checkout from 5 steps to 3."
  • Question every triplet. AI lists things in threes ("fast, reliable, and scalable"). Keep only items the source actually supports.
  • No synonym stacking. "Clear, concise, and actionable" → usually one word, often none.

Rhetorical tics

These create the feeling of content without content. Rule: if a sentence gets its effect from rhythm and contains no number, name or decision, make it concrete or delete it.

  • False contrast: "This isn't just a feature, it's a new way of working." Nobody claimed X; state Y directly. Keep the contrast only if someone actually argued X.
  • Staccato fragments: "No meetings. No friction. Just results." Turn into one concrete sentence or delete.
  • Fake suspense: "The result? 40% faster." / "Here's the thing:" State the answer.
  • Colon reveal: "The answer: simplicity."
  • Aphorism closers: "Because in the end, great products are built by great teams." Delete.
  • Reader instructions: "Let that sink in." / "Read that again."
  • "Whether you're X or Y…" openers.
  • Em-dash habit: a dash aside in every other sentence. Use commas, parentheses or a new sentence.

Formatting

  • One-sentence paragraphs stacked LinkedIn-style (outside of lists).
  • Bold sprinkled on random phrases. Bold only what a skimmer must not miss, if anything.
  • Bullets that are each a full paragraph; nesting deeper than two levels.
  • Emoji in headers.

Example

Before:

In today's rapidly evolving landscape, it is crucial that we leverage our existing capabilities to deliver a seamless and intuitive onboarding experience that empowers users to realize value quickly. This isn't just about reducing friction, it's about building lasting relationships.

After (source mentioned 60% day-one drop-off and a 7-screen flow):

60% of new users leave on day one. We're cutting onboarding from 7 screens to 3.

After (source mentioned neither):

We want to simplify onboarding. Open questions: What is the current day-one drop-off? What's the target? Which screens go?

The second "after" is the important case. Do not invent the 60%.

Report

After the rewritten doc, add a short block:

---
Unslop report
- Words: 2,840 → 610 (−79%)
- Decisions kept: 6
- Open questions raised: 5
- Invented content: none

Count words, never estimate them. If you can run code, use scripts/wordcount.py on the source and on your rewrite (python3 scripts/wordcount.py before.md after.md). If you can't, count by its rule: words of text only, markdown syntax (headers, table pipes, list markers, link URLs) excluded, and say the count is manual.

If you had to make any interpretation call (e.g. merging two contradictory scope statements), list it in one line each under the report. Nothing else: no summary of the doc, no list of every cut.

A full worked example (an iPad calculator PRD, before and after) is in the repo's examples/unslop.md.

Non-PRD inputs

For updates, memos, strategy docs or long messages, apply the same rules (extract, sentence test, gap check, slop pass, report) but skip the PRD template. Keep the doc's purpose and put the conclusion or ask in the first two sentences.

Don'ts

  • Don't add content, examples or metrics the source doesn't contain.
  • Don't soften open questions to be polite. A missing success metric is a real problem; say so plainly.
  • Don't preserve the source's section structure out of respect for it.
  • Don't end with a summary of what you did beyond the report.

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.

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.