ClawHub

Deslop CN

v1.0.0

当文章/文档/分享稿读起来像 AI 写的、充满套话、缺具体细节时使用。中英文反 AI 味两次 pass 重写。 不适用于代码注释和 commit message 的去 AI 味（用 slopbuster --mode code）。参见 doc-gen (用于生成文档)、human-writing (用于项目级中文...

⭐ 0· 12·0 current·0 all-time
by_silhouette@lanyasheng
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name, description, and SKILL.md consistently describe a tool for removing 'AI味' from prose (two-pass rewrite + scoring). The included pattern catalogs and tone guides directly support that purpose. There are no unrelated requirements (no env vars, binaries, or installs).
✓
Instruction Scope
Runtime instructions are explicit about workflow (diagnose, pass1, pass2, score), input (a file), and constraints (do not add new content or inject opinions). The skill does not instruct the agent to read unrelated system files, credentials, or external endpoints; it only operates on user-provided text. The SKILL.md includes clear output format and limits (e.g., not for academic papers or code comments).
✓
Install Mechanism
No install spec and no code files — this is instruction-only. Nothing will be written to disk or fetched during install. This minimizes installation risk.
✓
Credentials
The skill declares no required environment variables, no primary credential, and no config paths. The operations described (text diagnosis and rewrite) do not justify any extra secrets or system access, and none are requested.
✓
Persistence & Privilege
Flags show normal agent invocation (always: false). The skill does not request permanent presence or system-wide configuration changes, nor does it instruct modifying other skills. No elevated privileges are requested.
Assessment
This skill is internally consistent and low-risk from a supply-chain/permission perspective because it is instruction-only and asks for no credentials or installs. Things to consider before enabling: - Privacy: the skill will process any text you feed it. Do not submit secrets, private credentials, or regulated data unless you accept that the agent handling the skill will see that content. - Review outputs: the skill's workflow is explicit about not inventing new facts or opinions, but always review the rewritten text for accuracy, tone, and any unwanted changes. - Scope limits: the SKILL.md explicitly excludes academic papers and code/commit messages; use the alternative tools it references for those cases. - Runtime behavior: although the skill itself doesn't request network access or credentials, the agent runtime and model calls may transmit content to model APIs — ensure your environment's data-handling policies match your privacy needs. If you accept those conditions, the skill appears coherent with its stated purpose.

Like a lobster shell, security has layers — review code before you run it.

latestvk97dwdvf8tp1vswcfjexzg8fps84a0b9

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments