bg-szy/TOP-SKILLS / skills/agent-skills/avoid-ai-writing
avoid-ai-writing
VerifiedAudit and rewrite content to remove AI writing patterns ("AI-isms"). Supports detect-only audits, edit-in-place file cleanup, voice and context profiles, and an iterate-to-convergence pass.
| File | Kind | Size |
|---|---|---|
| CHANGELOG.md | markdown | 9.2 KB |
| CONTRIBUTING.md | markdown | 4.4 KB |
| GLOSSARY.md | markdown | 3.5 KB |
| LICENSE | other | 1.0 KB |
| LICENSE.txt | markdown | 1.0 KB |
| PROOF.md | markdown | 7.2 KB |
| README.md | markdown | 35.7 KB |
| SKILL.full.md | markdown | 151.4 KB |
| SKILL.md | skill md | 34.8 KB |
| package.json | manifest | 1.5 KB |
GLOSSARY.md
View on GitHub# Glossary
One-line definitions of the terms used across this repo's docs and issue templates. Each entry links to the canonical definition; when the two disagree, the linked source wins.
## Word tiers
| Term | Meaning |
|---|---|
| [Tier 1](references/patterns.md#tier-1--default-replacements) | Words to review on every match. Split into two bands, 1A and 1B, that use the same replacements but mean different things. |
| [Tier 1A](references/patterns.md#tier-1a--ai-frequency-markers) | AI frequency markers: words claimed to appear far more often in machine text. A cluster is evidence about how a passage was produced. |
| [Tier 1B](references/patterns.md#tier-1b--clarity-edits) | Clarity edits: wordiness and inflated formality. Not evidence of machine authorship. The detector emits them as `tier1-clarity`. |
| [Tier 2](references/patterns.md#tier-2--flag-when-2-appear-in-the-same-paragraph) | Words that are fine alone. Flagged when two or more appear in the same paragraph. |
| [Tier 3](references/patterns.md#tier-3--flag-only-at-high-density) | Common words flagged only at high density, roughly 3% or more of the text. |
| [Tier 3 phrases](references/patterns.md#tier-3-phrases--flag-at-density-or-in-clusters) | Multi-word boilerplate. Flagged when one phrase appears two or more times, or when three or more distinct phrases appear in one piece. |
## Severity
| Term | Meaning |
|---|---|
| [P0](SKILL.md#p0--credibility-killers-fix-immediately) | Credibility killers. Fix immediately. |
| [P1](SKILL.md#p1--obvious-ai-smell-fix-before-publishing) | Obvious AI smell. Fix before publishing. |
| [P2](SKILL.md#p2--stylistic-polish-fix-when-time-allows) | Stylistic polish. Fix when time allows. A quick pass covers P0 and P1; a full audit covers all three. |
## Modes and profiles
| Term | Meaning |
|---|---|
| [`rewrite` mode](SKILL.md#modes) | The default. Flags AI-isms and rewrites the text to fix them. |
| [`detect` mode](SKILL.md#modes) | Flags AI-isms only, with no rewriting. |
| [`edit` mode](SKILL.md#modes) | Edits a prose file in place with minimal, targeted changes. |
| [Context profile](references/patterns.md#context-profiles) | The audience hint (`linkedin`, `blog`, `technical-blog`, `investor-email`, `docs`, `casual`) that sets how strict each rule is. |
| [Voice profile](references/patterns.md#voice-profiles) | The optional persona (`casual`, `professional`, `technical`, `warm`, `blunt`) that sets how the prose should sound. Independent of the context profile. |
| [Register](https://github.com/conorbronsdon/avoid-ai-writing/blob/cc1d34319b4cc599bf503b80d35d7442bcde8047/.github/ISSUE_TEMPLATE/false_positive.yml) | The kind of writing: blog, docs, email, social post, and so on. The upstream false-positive report asks for it because a rule that holds in one register can misfire in another. |
| [Self-reference escape hatch](SKILL.md#self-reference-escape-hatch) | Quoted examples, code blocks, and text marked as illustrative are exempt from flagging, so writing *about* AI patterns is not scored for them. |
## Detector options
| Term | Meaning |
|---|---|
| [`contextMode`](detector/README.md#analyzetexttext-options--result) | The detector's context: `general` (default), `technical`, `marketing`, or `personal`. Each context profile maps to one; see [the mapping](references/patterns.md#detector-mode-mapping). |
| [`sourceMode`](detector/README.md#analyzetexttext-options--result) | How the detector reads input: `plain` (default) or `rendered-markdown`, which masks YAML frontmatter and HTML comments before scoring. |
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |