talk-normal
v0.6.1Stop LLM slop. A curated system prompt that cuts verbose, corporate-sounding LLM output by 56-71% (measured) while preserving information. Works bilingually...
⭐ 0· 25·0 current·0 all-time
by@hexiecs
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description promise (install a concise system prompt into agent config) matches the provided files (prompt.md) and the installer (install.sh). Nothing in the bundle requests unrelated capabilities (no network downloads, no extra binaries or secrets).
Instruction Scope
SKILL.md and prompt.md clearly state the skill installs rules into AGENTS.md and that behavior will be applied permanently until uninstalled. The installer script implements this exactly (inserts between BEGIN/END markers). Note: the skill's own metadata 'always: false' differs from the SKILL.md phrasing 'always-on behavior modifier' — that refers to the injected prompt being always-applied by the agent, not the skill being force-enabled platform-side.
Install Mechanism
No registry install or remote download. The provided install.sh is local, idempotent, and only edits or creates AGENTS.md. No archives or external URLs are fetched or executed.
Credentials
No environment variables, credentials, or config paths are required. The installer looks for AGENTS.md in '.', $HOME, or $OPENCLAW_WORKSPACE (reasonable for locating agent config).
Persistence & Privilege
The script intentionally injects persistent rules into AGENTS.md (so agent replies will follow them until removal). It does not set platform-level 'always' flags nor modify other skills. One operational caution: the installer searches $HOME and will create AGENTS.md in the current directory if none exists, which may result in unexpected changes if run from an unintended location—review target AGENTS.md before running.
Assessment
This bundle is coherent and lightweight, but review before running:
- Inspect prompt.md to confirm you agree with all rules that will be injected.
- Back up your existing AGENTS.md (if any) so you can restore it if the injected block isn't what you want.
- Run install.sh from the workspace where you expect AGENTS.md to live; the script searches '.', $HOME, and $OPENCLAW_WORKSPACE and will create ./AGENTS.md if none is found.
- Note the distinction: the skill makes your agent always apply these rules (by injecting them into AGENTS.md), but the skill itself is not a platform 'always: true' plugin. If you want to remove the rules later, run bash install.sh --uninstall.
- If you have doubts about provenance, verify the upstream GitHub repo and commit history before installing.Like a lobster shell, security has layers — review code before you run it.
latestvk97fa4f84n8knvk7f19ch47s6184ng7k
License
MIT-0
Free to use, modify, and redistribute. No attribution required.