内容写作优化技能

Fix AI writing patterns that create repetitive and robotic content

Security Scan

An instruction-only writing-style guide that asks for no credentials or installs and is consistent with its stated purpose, though package metadata (owner/slug/version) in _meta.json doesn't match the registry metadata and should be verified.

Purpose & Capability

The skill's name and description (fixing repetitive/robotic AI writing) match the SKILL.md content. There are no extra binaries, env vars, or config paths requested that would be unrelated to a writing-style guide. Note: _meta.json contains a different ownerId, slug, and version than the registry metadata — this mismatch is a packaging/metadata inconsistency worth verifying but does not change the skill's stated functionality.

Instruction Scope

SKILL.md contains only prose guidance and concrete style checks (e.g., paragraph openers, rhythm, nominalization). It does not instruct the agent to read files, access environment variables, call external endpoints, or perform system operations. The instructions are narrowly scoped to editing/writing behavior.

Install Mechanism

There is no install spec and no code files — the skill is instruction-only. That minimizes filesystem and supply-chain risk because nothing is downloaded or executed by the platform as part of installing this skill.

Credentials

The skill does not request any environment variables, credentials, or privileged config paths. Its operational needs are proportional to a text-editing guidance skill.

Persistence & Privilege

Flags are default (always: false, user-invocable: true, model invocation allowed). The skill does not request permanent presence or system-wide configuration changes.

Assessment

This appears to be a straightforward, instruction-only writing-style guide that does not ask for credentials or install code — generally low-risk. Before installing, verify the publisher/owner and the mismatch between the registry metadata and _meta.json (ownerId, slug, version) to ensure you have the intended package. Remember this skill will influence text the agent generates or edits; avoid feeding sensitive secrets or private data into prompts that the agent will process. If you require an auditable/trustworthy source, prefer a skill with a verifiable homepage, repository, or known publisher.

AI Writing Pattern Fixes

Paragraph Opener Trap

Scan first word of each paragraph - if 3+ start with "This/The/It" = robotic Vary with: "Meanwhile", "Consider", "Beyond that", action verbs, specific nouns Lead with action: "Marketing increased" not "The marketing team increased"

Rhythm Monotony Trap

Avoid same sentence length patterns - mix 5-word punches with 20-word flows Use strategic fragments. For emphasis. For pace breaks. Read aloud test: monotone rhythm = boring

Vague Claim Trap

Replace adjectives with numbers: "significant growth" → "40% growth" Swap evaluations for evidence: "impressive results" → "cut response time from 3s to 300ms" Delete weasel words: somewhat, fairly, quite, rather, slightly

List Overuse Trap

Use prose when explaining relationships or building arguments Use bullets only for instructions, features, options Flow test: if bullets feel choppy, try connected sentences

Parallel Structure Breaks

Match verb forms: "hiking, swimming, reading" not "hiking, swimming, to read" Match sentence patterns: "codes fast, debugs thoroughly, ships daily" Keep same grammatical structure in lists

Transition Word Crutches

Avoid: Furthermore, Moreover, Additionally, In conclusion Use echo technique: end paragraph with concept X, start next with concept X Pick up last idea and expand naturally

Voice Drift in Long Documents

Tone drift check: mark every 200 words - same person speaking? Pick 3-5 key phrases and use consistently throughout Don't switch from beginner-friendly to expert jargon mid-document

Word Economy Traps

Cut phrases: "in order to" → "to", "due to the fact that" → "because" One verb wins: "make improvement" → "improve", "conduct investigation" → "investigate" 80% rule: can you say it in 80% of words? Do it.

Nominalization Trap

Use verbs not nouns: "implementation of" → "implement" Action over abstraction: "decision-making process" → "decide" Zombie noun test: -tion, -ment, -ness endings often hide better verbs