360 Secure SkillHub
360 安全技能中心

audit system

该 Skill 提供一套纯指令式的结构化审计流程:界定范围、检查输入、识别问题、划分严重级别(Critical/High/Medium/Low)、给出具体修复与替代方案,并按固定模板输出审计报告,适用于代码、工作流、产品体验和提示词/智能体场景。它只分析所提供的内容,不执行外部验证。

mededdahby
mededdahby
热度 85.67
最近更新 2026-05-11 17:35:07
分类:开发依赖:纯提示词应用场景:代码审查 / 调试排错 / 报告撰写 / 工作流

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add audit-system
Metadata
name
audit-system
description
Perform structured audits on code, workflows, prompts, and products. Use when: (1) Something is not working as expected, (2) User asks for review or feedback, (3) A system behaves inconsistently, (4) Before deploying or shipping, (5) When optimizing performance or UX, (6) When debugging recurring issues.

Audit System Skill

Perform structured audits and generate actionable reports with clear severity, evidence, and fixes.

This is an instruction-only skill. It does not perform external verification, blockchain auditing, or legal certification.


Quick Reference

Situation Action
Code not working Run Code Audit
Workflow failing Run Workflow Audit
UX feels bad Run Product Audit
Prompt/AI unstable Run Prompt Audit
Before deploy Run Full Audit
Repeated bugs Focus on root-cause analysis

Audit Types

1. Code Audit

Check:

  • logic errors
  • missing validation
  • security risks
  • bad patterns
  • performance issues

2. Workflow Audit

Check:

  • broken steps
  • missing retries
  • failure points
  • unnecessary complexity
  • automation gaps

3. Product Audit

Check:

  • onboarding friction
  • unclear UX
  • conversion blockers
  • trust issues
  • missing features

4. Prompt / Agent Audit

Check:

  • unclear instructions
  • conflicting rules
  • missing constraints
  • unstable outputs
  • over-autonomy risks

Audit Process

Step 1 — Define Scope

Identify:

  • what is being audited
  • expected behavior
  • actual behavior
  • available data

Step 2 — Inspect

Analyze inputs:

  • code
  • prompts
  • configs
  • logs
  • workflows

Look for:

  • inconsistencies
  • missing logic
  • unclear flow
  • hidden risks

Step 3 — Detect Issues

For each issue:

  • describe clearly
  • link to evidence
  • explain impact

Step 4 — Classify Severity

  • Critical → breaks system / risk of loss
  • High → likely failure
  • Medium → important weakness
  • Low → improvement

Step 5 — Recommend Fixes

For each issue:

  • what to fix
  • why it matters
  • exact fix
  • quick workaround

Step 6 — Prioritize

Always output:

  • top 3 issues
  • quick wins
  • long-term fixes

Output Format

Audit Report

Scope

  • Target:
  • Type:
  • Evidence:
  • Limitations:

Findings

[Severity] Title

  • Area:
  • Problem:
  • Evidence:
  • Impact:
  • Fix:

Priority Actions

  1. ...
  2. ...
  3. ...

Quick Wins

  • ...
  • ...

Long-Term Improvements

  • ...

Open Questions

  • ...

Behavior Rules

  • Be precise, not vague
  • Do not invent missing data
  • Do not exaggerate severity
  • Do not claim certification
  • Focus on actionable fixes

When NOT to use this skill

Do NOT use for:

  • legal certification
  • financial compliance guarantees
  • blockchain verification
  • cryptographic proof generation

Only analyze what is provided.


Upgrade Path (Advanced)

If repeated issues appear:

  • suggest system redesign
  • suggest automation improvements
  • suggest monitoring/logging additions