agentleFS
geronimo-iiageronimo-iia/agent-skills/skills/anti-slop/SKILL.md

anti-slop

Skill2 starsChanged 3 months agoMITView on GitHub

Strip AI slop from written artifacts: KB documents, analysis, synthesis, research notes, technical docs, PR descriptions. Enforces structural writing rules and blocks known slop patterns. Invoke before writing any non-trivial prose artifact. Not for chat responses (caveman mode covers those).

31 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use research and handoff.

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.
  • NVIDIASkill

    research

    Before designing a feature, research how comparable libraries solve the problem, check best practices, known pitfalls, and upstream changes. Use when starting a new feature, investigating an unfamiliar area, or the user asks to research something.

    Also in 3 projects that use this

  • MicrosoftSkill

    handoff

    Compact the current conversation into a handoff document for another agent to pick up.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 30 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never restate what was just written.

    line 24
  2. Distinctive:

    Every word earns its place: Read the draft. Delete any sentence that could be removed without losing information.

    line 32
  3. Very specific:

    No --- horizontal rules: Never use --- as a section separator in documents or responses.

    line 86
  4. Distinctive:

    Headers earn their place: Use H2/H3 only when the section is long enough to warrant navigation.

    line 89
  5. Distinctive:

    No 3-level bullet nesting: If bullets nest 3 deep, collapse to prose or restructure.

    line 92
  6. Distinctive:

    No single-caller abstractions: No interface, wrapper, or helper with one caller unless re-use is certain.

    line 118
  7. Distinctive:

    Run before finalizing any written artifact:

    line 130

“Common” is measured: kept by at least 6 in 10 of the 30 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

Anti-Slop Writing Rules

Apply before writing any KB document, synthesis, analysis, report, or technical documentation. These rules govern written artifacts, not chat.

Structural rules

Start with content. First sentence delivers the fact, claim, or decision. No "In this document we will explore...", no "This analysis covers...", no "It is worth noting that...".

No concluding recap. Final paragraph must add new information or be absent. Never restate what was just written.

Active voice. "The algorithm reduces latency" not "Latency is reduced by the algorithm."

One idea per sentence. Split compound sentences. Short is readable; long is a symptom.

Every word earns its place. Read the draft. Delete any sentence that could be removed without losing information.

Slop patterns — detect and cut

Confidence theater

Hedging on things that are simply true or false.

  • Wrong: "This approach may potentially reduce overhead in certain scenarios."
  • Right: "This approach reduces overhead when the input is sorted."

Fix: state the fact directly. If genuinely uncertain, name the condition (when X, if Y, assuming Z) — do not hedge with vague qualifiers.

False enumeration

Turning a 2-item answer into a 5-bullet list for appearance of thoroughness.

  • Wrong: five bullets when three are filler restating the same point
  • Right: list only when items are genuinely parallel and enumeration adds scannability. Prose is often clearer for 2–3 items.

Fix: ask "would this be clearer as one sentence?" If yes, write one sentence.

Balance theater

Adding a counterpoint to seem fair when one position is correct.

  • Wrong: "While approach A has merits, approach B also has its advantages..." (when A is clearly better for the stated goal)
  • Right: state the correct position. If a real tradeoff exists, name it specifically with the exact condition under which it matters.

Fix: only acknowledge opposing positions when they apply to the concrete situation. Generic balance is noise.

Fake precision

Claiming authority without a traceable source.

  • Wrong: "Studies show...", "Many experts agree...", "Research indicates..."
  • Right: name the source or drop the claim. Use [DIRECT] / [INFERRED] / [SPECULATIVE] markers from kb-conventions for KB documents.

Banned words

Cut on sight: delve, leverage, utilize, robust, seamlessly, tapestry, testament, spearhead, in conclusion, furthermore, moreover, it's worth noting, it is important to note, please note, rapidly evolving, it's important to remember.

Formatting rules

No --- horizontal rules. Never use --- as a section separator in documents or responses. Frontmatter delimiters only.

Headers earn their place. Use H2/H3 only when the section is long enough to warrant navigation. One-paragraph sections need prose, not a header.

No 3-level bullet nesting. If bullets nest 3 deep, collapse to prose or restructure. Deep nesting signals the content isn't actually parallel.

Bold means emphasis, not decoration. Bold one or two genuinely critical terms per document. Bolding every other phrase means nothing is emphasized.

No code ticks on non-code. Ticks for commands, identifiers, file paths, values — not for concepts or invented terms that sound more technical in ticks.

Tables require comparison value. A table is justified when the reader needs to scan across columns (comparison, lookup). Three items with a label and description are a list, not a table.

No padding whitespace. Blank line between paragraphs — yes. Blank line between every sentence or bullet — no. Whitespace structures; it does not signal thoroughness.

No concept inflation. "the pipeline", "the architecture", "the system" — lowercase unless a proper name. Unnecessary capitalization implies official status that doesn't exist.

Code artifact rules

No restate comments. Comment only the why — hidden constraint, non-obvious invariant, workaround for a specific bug. Never describe what the code does.

No single-caller abstractions. No interface, wrapper, or helper with one caller unless re-use is certain.

No TODO placeholders unless explicitly requested.

Slop audit before submitting:

  • Does any dependency exist only to replace a native 5-line solution?
  • Is any abstraction layer present that the caller never benefits from?
  • Are there any "just in case" parameters or flags not used by existing callers?

Self-audit checklist

Run before finalizing any written artifact:

  • First sentence delivers content, no intro
  • Last paragraph adds new information, not recap
  • Every list has genuinely parallel, enumerable items
  • Every hedge names a concrete condition
  • Every authority claim has a traceable source
  • No banned words present
  • Active voice throughout
  • No --- separators outside frontmatter
  • No headers on single-paragraph sections
  • No bold used decoratively
  • No tables where a list or sentence suffices
  • No concept inflation (unnecessary capitalization)

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.