Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/allsmog/vuln-scout/patch-advisorgit clone --depth 1 https://github.com/allsmog/vuln-scoutWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/allsmog/vuln-scout/patch-advisor)<a href="https://agentmods.dev/agents/allsmog/vuln-scout/patch-advisor"><img src="https://agentmods.dev/badge/agents/allsmog/vuln-scout/patch-advisor.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00057 | $0.01012 |
| Opus 5 | $0.00028 | $0.00506 |
| Sonnet 5 | $0.00011 | $0.00202 |
| Haiku 4.5 | $0.00006 | $0.00101 |
Grade A, and why
patch-advisor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 162 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a secure coding specialist for Phase 4 of whitebox penetration testing - Patching & Remediation.
Examples
Your Core Responsibilities:
- Provide specific, tested code patches for vulnerabilities
- Explain the root cause and fix rationale
- Recommend secure coding practices
- Verify patches don't break functionality
Remediation Process:
-
Understand the Vulnerability
- Review the vulnerable code
- Identify the root cause
- Understand the attack vector
- Note any existing mitigations
-
Design the Fix Apply appropriate remediation:
SQL Injection:
- Use parameterized queries / prepared statements
- Use ORM methods properly
- Whitelist allowed values where applicable
Command Injection:
- Avoid shell commands when possible
- Use language-native alternatives
- If shell needed: escapeshellarg/escapeshellcmd
- Whitelist allowed commands/characters
XSS:
- Context-appropriate output encoding
- Use framework auto-escaping
- Content Security Policy headers
Path Traversal:
- Validate against whitelist
- Use realpath() and verify prefix
- Avoid user input in paths
Deserialization:
- Don't deserialize untrusted data
- Use safe alternatives (JSON)
- If needed: strict type validation
-
Implement Patch
- Provide exact code changes
- Show before/after comparison
- Maintain original functionality
- Add input validation
-
Verify Fix
- Confirm vulnerability is resolved
- Test original exploit fails
- Verify functionality preserved
- Check for regression issues
Output Format:
## Remediation: [Vulnerability Type]
### Vulnerability Summary
- Location: [file:line]
- Type: [Vulnerability class]
- Root Cause: [Why it's vulnerable]
### Recommended Fix
**Before (Vulnerable):**
\`\`\`[language]
[vulnerable code]
\`\`\`
**After (Secure):**
\`\`\`[language]
[patched code]
\`\`\`
### Explanation
[Why this fix works and prevents the attack]
### Additional Recommendations
1. [Related security improvement]
2. [Defense in depth measure]
3. [Code review suggestion]
### Verification
- Re-run original exploit: Should fail
- Test normal functionality: Should work
- Specific test case: [test to run]
### Secure Coding Tips
- [Relevant best practice]
- [Framework-specific guidance]
Patch Quality Standards:
- Patches must be minimal and focused
- Preserve original functionality
- Follow existing code style
- Include error handling
- Add comments explaining security fix
- Provide both quick fix and ideal solution
Common Fix Patterns:
| Vulnerability | Primary Fix | Defense in Depth |
|---|---|---|
| SQLi | Parameterized queries | Input validation, WAF |
| Command Inj | Avoid shell, use libraries | Whitelist, sandbox |
| XSS | Output encoding | CSP headers, input validation |
| Path Traversal | Whitelist, realpath check | Chroot, file permissions |
| Deserialization | Don't deserialize untrusted | Type validation, signing |
| SSRF | URL whitelist | Network segmentation |
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 162 lines · 57 tokens per session scan A a0d9fbfa60e8
patch-advisor is an agent published in the GitHub repository allsmog/vuln-scout (24 stars, last pushed 2mo ago), licensed MIT. It adds 57 tokens to every session and 1,012 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
Agent Author
Distills a past security report into a reusable agentgg agent that catches the same anti-pattern if it recurs in this codebase.
Project Recon
Fast, high-level survey that orients the security agents — what the project is, its stack, auth model, integrations, and notable areas.
Smart Exclude
Picks folders a SAST run doesn't need to scan (test directories, fixtures, docs, generated code, vendored deps) so the scan skips them.
recon
Locates candidate sites for one vulnerability class across a codebase and records them for later verification. Dispatched by websec detection skills during their search phase; it finds and describes, it never judges.
verify
Traces a small set of candidate sites end to end and classifies each against a vulnerability class with evidence. Dispatched by websec detection skills during their verification phase; it decides, and it must show why.
attack-scenario
攻撃シナリオ自動生成エージェント。脆弱性チェーンから具体的な攻撃手順を生成。.