AgentSkillsCN

confident-language-guard

在撰写或更新文档中的各类指令时使用（例如：ROADMAP、AGENTS.md、CLAUDE.md、skills），作为温和的提醒，帮助您保持语言措辞谨慎而灵活；除非确有需要，否则应避免使用过于自信或绝对化的指令。

SKILL.md
--- frontmatter
name: confident-language-guard
description: Use when drafting or updating directives in docs (e.g., ROADMAP, AGENTS.md, CLAUDE.md, skills) as a gentle reminder to keep language cautious and flexible; avoid overly confident or absolute directives unless explicitly required.

Confident Language Guard

Goal

Keep documentation helpful without overstating certainty or locking in fragile guidance. This is a reminder, not an iron-clad rule.

Core Rules

  • •Prefer scoped, time-bound statements ("in this phase", "for the current repo state").
  • •Replace absolutes ("always", "never", "must") with softer language unless a hard rule truly exists.
  • •Use qualifiers when the information could change ("appears", "likely", "based on current files").
  • •Separate recommendations from requirements and label them clearly.
  • •State assumptions when they affect the guidance.
  • •Avoid turning personal judgment into policy language.

Allowed Absolutes (Exceptions)

  • •Use absolute language only when a requirement is explicit (e.g., system or repo rules).
  • •If a hard rule exists, cite the source or file when possible.

Quick Checklist (Docs Pass)

  • •Does every directive have a clear source or rationale?
  • •Can any "always/never/must" be softened without losing meaning?
  • •Are assumptions stated and scoped?
  • •Are recommendations labeled as such?

Example Rewrite

  • •Before: "You must always run full tests before every commit."
  • •After: "It is generally safer to run the full test suite before committing; skip only when time is constrained or the change is purely textual."

[Codex - 2026-01-12]