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#

SituationActionCode not workingRun Code AuditWorkflow failingRun Workflow AuditUX feels badRun Product AuditPrompt/AI unstableRun Prompt AuditBefore deployRun Full AuditRepeated bugsFocus 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