Click to expand
This is a Codex Skill repo. It gives your coding assistant a repeatable workflow instead of one-off prompts.
| Challenge | Without this skill | With this skill |
|---|---|---|
| Getting started | You re-explain context every time | The workflow is encoded in SKILL.md |
| Output quality | Inconsistent formatting | Consistent outputs + guardrails |
| Scope control | Easy to drift | Clear stop/confirm gates |
| Reuse | Hard to share | Share it as a skill folder |
| Speed | Manual, repetitive steps | Short prompts trigger full workflows |
- ✅ You want a stable workflow in Codex/Claude/Gemini CLI skills.
- ✅ You want consistent deliverables (tables/diagrams/docs/reports).
- ✅ You prefer “ask for inputs first, then generate”.
- ❌ One-off work where you do not want workflow constraints.
- ❌ Scenarios that require silent assumptions (these skills avoid guessing).
审计"、"审查"、"验证"、"检查质量"、"分析提示词"、"评估技能"、
- Put this folder into your Codex skills directory:
Windows: %USERPROFILE%\.codex\skills\hello-auditkit
macOS/Linux: ~/.codex/skills/hello-auditkit
- Restart your CLI tool (if needed) so it reloads skills.
- In chat, mention
hello-auditkitor use a trigger keyword.
-
Say:
audit this SKILL.md(or provide a path). -
It writes the audit report to a file.
-
Review, then decide whether to apply fixes.
📊 Click to view the workflow
flowchart TD
A[User prompt] --> B{Trigger match}
B -->|match| C[Load SKILL.md + references/]
C --> D[Ask for missing inputs]
D --> E[Generate outputs]
B -->|no match| F[Do nothing]
style A fill:#e3f2fd
style C fill:#fff3e0
style E fill:#4caf50,color:#fff
style F fill:#ffebee
- Start here:
SKILL.md - If present:
references/andassets/(loaded on-demand)
Q: Does it modify files automatically?
A: No. It requires explicit confirmation before applying any fix.
Q: Where is the audit report?
A: It is saved to a file (terminal output stays minimal).
Q: Can I audit a folder?
A: Yes, provide a folder path and specify what to check.
Q: Does it use the web?
A: It may fetch the latest prompting guidelines if needed.
Q: What if the target is unclear?
A: It will ask 1–3 clarification questions.
Q: Can I limit the scope?
A: Yes, tell it exactly which files or checks to run.
Problem: You did not provide a valid path/target.
Fix: Provide a concrete file/folder path and retry.
Problem: The audit goal is ambiguous.
Fix: Reply with the exact target and desired depth.
Problem: The target file changed after reading.
Fix: Re-run after saving the latest file.
Problem: Network access is unavailable.
Fix: Run offline mode or enable network.
Problem: Host tool settings may differ.
Fix: Ask it to keep terminal output minimal and write to files.
- Current version:
2.0.0(seeSKILL.md)
- Do not paste secrets/tokens into prompts or sample files.
- If you find a security issue, open a minimal report without sensitive details.
See CONTRIBUTING.md.
This repository is licensed under MIT. See LICENSE.
Made by hellowind777