Audit Extractor
Documentsaudit-extractorAudit Extractor is an Agent Skill for documents published in the Layr-Labs/eigenlayer-contracts repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Extract findings from PDF audit reports and convert them to a markdown checklist. Use when the user asks to process an audit report, extract audit findings, or create an audit action items list. (project)
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/Layr-Labs/eigenlayer-contracts/blob/HEAD/.claude/skills/audit-extractor/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/audit-extractor, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Audit Extractor
Extract findings from PDF security audit reports and convert them to a structured markdown checklist for tracking remediation.
Overview
This skill processes PDF audit reports (from firms like Certora, Sigma Prime, Cantina, etc.) and extracts:
- Finding ID (e.g., C-01, H-01, M-01, L-01, I-01)
- Title/Description
- Severity (Critical, High, Medium, Low, Informational)
- Status (Fixed, Partially fixed, Awaiting Fix, Acknowledged, Disregarded, Pending)
- Impact and Likelihood (when available)
- Detailed descriptions and recommendations
How to Extract Findings
Step 1: Read the PDF directly
Use the Read tool to read the PDF file. Claude can natively read and understand PDF documents:
Read the PDF file at: <pdf_path>
Step 2: Parse the audit content
After reading the PDF, manually extract findings by looking for:
-
Summary/Findings Table - Usually contains:
- Finding IDs (patterns like C-01, H-01, M-01, L-01, I-01)
- Titles
- Severity levels
- Status
-
Detailed Finding Sections - Usually organized by PR or category, containing:
- Full descriptions
- Code snippets
- Exploit scenarios
- Recommendations
- Customer response / Fix review status
-
Severity Information - Look for:
- Impact (High, Medium, Low)
- Likelihood (High, Medium, Low)
Step 3: Generate the markdown output
Create a markdown file with the following structure:
# Audit Findings: [Audit Name]
**Auditor:** [Auditor Name]
**Date:** [Date]
**Status:** [Draft/Final]
**Total findings:** [Count]
## Summary
| Severity | Discovered | Confirmed | Fixed |
|----------|------------|-----------|-------|
| Critical | X | | |
| High | X | | |
| Medium | X | | |
| Low | X | | |
| Informational | X | | |
| **Total** | **X** | | |
## Action Required (X items pending)
- [ ] **M-01** (Medium): [Title]
- [ ] **L-01** (Low): [Title]
---
## All Findings
### ([PR/Category Name])
#### [ID]. [Title]
| Severity | Impact | Likelihood | Status |
|----------|--------|------------|--------|
| [Severity] | [Impact] | [Likelihood] | [Status] |
**Files:** `[filename.sol]`
**Description:** [Full description]
**Recommendation:** [Recommendation text]
---
Step 4: Save the output
Use the Write tool to save the markdown file:
- Save to
audits/[Audit-Name]-Findings.md
Output Format Guidelines
Severity Levels
- Critical - Direct loss of funds or complete protocol compromise
- High - Significant impact on protocol security or functionality
- Medium - Moderate impact, potential for exploitation under certain conditions
- Low - Minor issues, best practice violations
- Informational - Code quality, gas optimizations, documentation
Status Values
- Fixed - Issue has been resolved
- Partially fixed - Issue has been partially addressed
- Awaiting Fix - Issue acknowledged, fix pending
- Acknowledged - Issue acknowledged, may not be fixed
- Pending - Awaiting customer response
- Disregarded - Issue will not be fixed (by design)
Checklist Rules
- Use
[ ]for unresolved items (Pending, Awaiting Fix, Partially fixed) - Use
[x]for resolved items (Fixed, Acknowledged, Disregarded)
Audit Files Location
Audit reports are typically stored in: audits/
Example Extraction
Given a PDF with findings like:
M-01 MigrateSlashers may assign incompatible slasher | Medium | Pending
L-01 Instant slasher setting leaves stale field | Low | Pending
I-01 getSlasher is implemented twice | Informational | Pending
Generate:
# Audit Findings: EigenLayer - Slashing UX Improvements
**Auditor:** Certora
**Date:** December 2025
**Total findings:** 3
## Summary
| Severity | Count |
|----------|-------|
| Medium | 1 |
| Low | 1 |
| Informational | 1 |
## Action Required (2 items)
- [ ] **M-01** (Medium): MigrateSlashers may assign incompatible slasher
- [ ] **L-01** (Low): Instant slasher setting leaves stale field
## All Findings
### Medium (1)
#### M-01. MigrateSlashers may assign an incompatible slasher address
| Severity | Impact | Likelihood | Status |
|----------|--------|------------|--------|
| Medium | High | Low | Pending |
**Description:** [Full description from PDF]
**Recommendation:** [Recommendation from PDF]
---
### Low (1)
#### L-01. Instant slasher setting leaves stale slasher field in storage
| Severity | Impact | Likelihood | Status |
|----------|--------|------------|--------|
| Low | Low | Low | Pending |
**Description:** [Full description from PDF]
**Recommendation:** [Recommendation from PDF]
---
### Informational (1)
#### I-01. getSlasher is implemented twice
| Severity | Status |
|----------|--------|
| Informational | Pending |
**Description:** [Full description from PDF]
**Recommendation:** [Recommendation from PDF]
Troubleshooting
PDF won't read
- Ensure the file path is correct and the file exists
- Check if the PDF is password protected (not supported)
Missing findings
- Some PDFs have findings spread across multiple sections
- Check the Table of Contents for finding locations
- Look for "Detailed Findings" or similar sections
Inconsistent formatting
- Different audit firms use different formats
- Adapt the markdown structure to match the source format
- Preserve all relevant information even if structure differs