/auto-fix
[experimental] Auto-remediate verified findings by generating patches and optionally creating a PR
$ npx -y skills add allsmog/vuln-scout --agent claude-codeHow it fires
How this command gets triggered: by you, by Claude, or both.
- Fires itselfClaude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/auto-fix
Context preview
What this command does when you run it.
[experimental] Auto-remediate verified findings by generating patches and optionally creating a PR
Command definition
auto-fix.mdname: auto-fix
description: "[experimental] Auto-remediate verified findings by generating patches and optionally creating a PR"
argument-hint: "[--finding VSCOUT-XXXX] [--all-verified] [--severity critical,high] [--dry-run] [--create-pr]"
allowed-tools:
- Bash
- Glob
- Grep
- Read
- Write
- Task
- AskUserQuestion
Auto-Fix Security Findings
Automatically generate and apply patches for verified security findings. Uses the `patch-advisor` agent to create context-aware fixes.
Usage
/vuln-scout:auto-fix --finding VSCOUT-0003
/vuln-scout:auto-fix --all-verified --severity critical,high
/vuln-scout:auto-fix --all-verified --dry-run
/vuln-scout:auto-fix --all-verified --create-pr
Flags
| Flag | Effect | |------|--------| | `--finding` | Fix a specific finding by ID | | `--all-verified` | Fix all verified findings | | `--severity` | Filter to specific severities (comma-separated) | | `--dry-run` | Show proposed patches without applying | | `--create-pr` | Create a git branch and open a PR with the fixes |
Safety
- Only operates on `verdict: verified` findings (never patches unverified code)
- Always shows the proposed change before applying (unless `--create-pr` in batch mode)
- Never auto-merges PRs -- always requires human review
- Re-scans after patching to verify the fix eliminated the finding
Workflow
Step 1: Load findings
Read `.claude/findings.json`. Filter to `verdict: "verified"` findings matching the specified criteria.
Step 2: Generate fixes
For each selected finding: 1. Read the vulnerable code context (file + surrounding lines) 2. Use the `patch-advisor` agent's remediation knowledge to generate a fix 3. Show the before/after diff to the user
Step 3: Apply patches
If not `--dry-run`: 1. Apply the code change using the Write tool 2. Re-run scan on the modified file to verify the finding is resolved 3. If the finding persists, warn the user and revert
Step 4: Create PR (if `--create-pr`)
BRANCH="vuln-scout/auto-fix-$(date +%Y%m%d-%H%M%S)"
git checkout -b "$BRANCH"
git add <modified-files>
git commit -m "fix(security): remediate <finding-ids> [vuln-scout auto-fix]"
gh pr create --title "Security: Auto-fix <N> findings" --body "<details>"
Read more
name: auto-fix description: "[experimental] Auto-remediate verified findings by generating patches and optionally creating a PR" argument-hint: "[--finding VSCOUT-XXXX] [--all-verified] [--severity critical,high] [--dry-run] [--create-pr]" allowed-tools: - Bash - Glob - Grep - Read - Write - Task - AskUserQuestion
Auto-Fix Security Findings
Automatically generate and apply patches for verified security findings. Uses the `patch-advisor` agent to create context-aware fixes.
Usage
/vuln-scout:auto-fix --finding VSCOUT-0003 /vuln-scout:auto-fix --all-verified --severity critical,high /vuln-scout:auto-fix --all-verified --dry-run /vuln-scout:auto-fix --all-verified --create-pr
Flags
| Flag | Effect | |------|--------| | `--finding` | Fix a specific finding by ID | | `--all-verified` | Fix all verified findings | | `--severity` | Filter to specific severities (comma-separated) | | `--dry-run` | Show proposed patches without applying | | `--create-pr` | Create a git branch and open a PR with the fixes |
Safety
- Only operates on `verdict: verified` findings (never patches unverified code)
- Always shows the proposed change before applying (unless `--create-pr` in batch mode)
- Never auto-merges PRs -- always requires human review
- Re-scans after patching to verify the fix eliminated the finding
Workflow
Step 1: Load findings
Read `.claude/findings.json`. Filter to `verdict: "verified"` findings matching the specified criteria.
Step 2: Generate fixes
For each selected finding: 1. Read the vulnerable code context (file + surrounding lines) 2. Use the `patch-advisor` agent's remediation knowledge to generate a fix 3. Show the before/after diff to the user
Step 3: Apply patches
If not `--dry-run`: 1. Apply the code change using the Write tool 2. Re-run scan on the modified file to verify the finding is resolved 3. If the finding persists, warn the user and revert
Step 4: Create PR (if `--create-pr`)
BRANCH="vuln-scout/auto-fix-$(date +%Y%m%d-%H%M%S)" git checkout -b "$BRANCH" git add <modified-files> git commit -m "fix(security): remediate <finding-ids> [vuln-scout auto-fix]" gh pr create --title "Security: Auto-fix <N> findings" --body "<details>"
AI-powered whitebox penetration testing plugin for Claude Code. 9 languages, 22 skills, 7 autonomous agents. STRIDE threat modeling, OWASP 2025 coverage, polyglot monorepo support.
Repo: allsmog/vuln-scout
Other commands on vuln-scout.
- /create-rule
[experimental] Create a custom Semgrep detection rule from a confirmed vulnerability pattern
Open command - /diff
[stable] Compare security posture between two git refs to find new/fixed vulnerabilities and track regression
Open command - /full-audit
[stable] End-to-end security audit with hotspot-aware framework pivots, shared findings.json schema, and CI-friendly workflow flags
Open command - /mobile-audit
[beta] Audit a decompiled Android target — scans jadx_out/sources + apktool_out together and merges findings
Open command - /mutate
[experimental] Security mutation testing -- weaken security controls and check if the scanner detects the resulting vulnerability
Open command - /org-memory-compile
[experimental] Compile human-reviewed scan history into local organization memory
Open command

