validate — Claude Code command: filter findings before submission
validate — a validation command that culls weak findings before report writing. Runs a 7-Question Gate and 4-gate checklist to prevent submissions that damage your validity ratio.
MIT ★ 16,193 updated 2026-10-29
/plugin marketplace add elementalsouls/Claude-BugHunter
/plugin install claude-bughunter@elementalsoulsWhat validate does
The command walks through a structured set of questions and checks designed to reject findings that are not ready for submission. The 7-Question Gate asks whether you can reproduce the vulnerability right now, whether it falls within program scope, whether it requires admin privileges an attacker cannot obtain, whether it's already known, whether the impact goes beyond technical possibility to real compromise, and whether it lands on the never-submit list. Followed by 4 gates: Gate 0 confirms real HTTP requests and reproducibility; Gate 1 verifies actual impact (not just data visibility); Gate 2 searches for duplicates in disclosed reports and changelogs; Gate 3 checks report quality against a title formula and PoC standards.
The output is PASS (write the report), KILL (move on, with the failed question and reason), or DOWNGRADE (findings is valid but impact proof is weak).
What it helps with
- reject findings that the program won't accept or that are already reported
- confirm impact through real evidence, not hypothetical scenarios
- verify the vulnerability touches an in-scope asset, not staging or third-party services
- filter out self-XSS, open redirects without a chain, and other low-value bug classes
- assess report quality before investing time in writing
Example prompts for the agent
- /validate
- I found an IDOR on /users/{id}/profile. /validate to check if it's worth reporting.
How to install validate
1Plugin marketplace
/plugin marketplace add elementalsouls/Claude-BugHunter/plugin install claude-bughunter@elementalsouls2Manually: put the files in place
.claude/commands/validate.mdCall it with /validate. For all projects use ~/.claude/commands/
Manually: put the files in place
~/.codex/prompts/validate.mdCall it with /prompts:validate. Codex marks this format deprecated in favour of skills
Manually: put the files in place
.github/prompts/validate.prompt.mdManually: put the files in place
.gemini/commands/validate.tomlGemini stores commands as TOML: the prompt text goes into the prompt field
Questions about validate
How do I install validate in Claude Code?
Run /plugin marketplace add elementalsouls/Claude-BugHunter, then /plugin install claude-bughunter@elementalsouls, then add .claude/commands/validate.md to your project.
What does validate do?
It runs a finding through a 7-Question Gate and 4 checklists before submission. It stops weak findings at validation so you don't waste time writing reports the program will reject.
What happens if a finding fails one question?
The command outputs KILL and tells you which question failed and why. One wrong answer means the finding doesn't move forward. If impact is only technically possible but not proven, it outputs DOWNGRADE instead, asking for stronger proof.
Source text
validate.md · MIT
--- name: validate description: Validate a finding — runs 7-Question Gate + 4-gate checklist. Kills weak findings before report writing. Prevents N/A submissions that hurt validity ratio. Usage: /validate --- # /validate Run full validation on the current finding before writing a report. ## What This Does 1. Runs 7-Question Gate (one wrong answer = kill it) 2. Checks against the always-rejected list 3. Runs 4 pre-submission gates 4. Outputs: PASS (write the report) or KILL (move on) ## Usage ``` /validate ``` Describe the finding when prompted. Include: - The endpoint - The bug class - What the PoC shows - The target program ## The 7-Question Gate Answer each. ONE wrong answer = STOP. ### Q1: Can I demonstrate this step-by-step RIGHT NOW? Write this out: ``` 1. Setup: I need [own account / another user's ID / no account] 2. Request: [exact HTTP method, URL, headers, body] 3. Result: Response shows [exact data / action completed] 4. Impact: Real consequence is [account takeover / PII exposed / money stolen] 5. Cost: Time: [X min], Capital: [$0 / $X] ``` If step 2 is "I need to look at the code more" → KILL IT. ### Q2: Is the impact accepted by this program? Check program scope page. Is your bug class listed? Is it excluded? ### Q3: Is the vulnerable asset in scope? Exact domain in scope? Not staging/dev? Not a third-party service? ### Q4: Does it need admin or privileged access that an attacker can't get? "Admin can do X" → KILL IT. "Regular user can do X that only admin should" → valid. ### Q5: Is this known or documented behavior? Search disclosed reports + changelog + API docs. ### Q6: Can you prove impact beyond "technically possible"? - XSS → actual cookie value in exfil request, not just alert() - SSRF → response body from internal service, not just DNS callback - IDOR → actual other-user's private data in response, not just 200 status ### Q7: Is this on the never-submit list? ``` Missing headers, GraphQL introspection alone, clickjacking without PoC, self-XSS, open redirect alone, SSRF DNS-only, logout CSRF, banner disclosure, rate limit on non-critical forms, missing cookie flags alone... ``` If yes → KILL IT unless you have a chain. ## Check: Conditionally Valid? If it's on the never-submit list, can you chain it? | You Have | Chain Available? | |---|---| | Open redirect | + OAuth code theft → ATO? | | SSRF DNS-only | + internal service data? | | Clickjacking | + sensitive action + PoC? | | CORS wildcard | + credentialed data exfil? | | Prompt injection | + IDOR → other user's data? | If no chain → KILL IT. If chain confirmed → report both together. ## 4 Gates — All Must Pass **Gate 0 (30 sec):** ``` [ ] Confirmed with real HTTP requests (not just code reading) [ ] In scope (checked program page) [ ] Reproducible from scratch [ ] Evidence captured ``` **Gate 1 — Impact (2 min):** ``` [ ] Can answer "What does attacker walk away with?" [ ] More than "sees non-sensitive data" [ ] Real victim exists [ ] No unlikely preconditions ``` **Gate 2 — Dedup (5 min):** ``` [ ] Searched HackerOne Hacktivity for endpoint + bug class [ ] Searched GitHub issues [ ] Read 5 most recent disclosed reports [ ] Not in changelog as known issue ``` **Gate 3 — Report quality (10 min):** ``` [ ] Title formula: [Class] in [Endpoint] allows [actor] to [impact] [ ] Steps have exact HTTP request [ ] Evidence shows actual impact [ ] CVSS calculated [ ] Fix: 1-2 concrete sentences ``` ## Output **PASS:** "All 7 questions pass. All 4 gates pass. Proceed to /report." **KILL:** "Q[N] fails because [reason]. Kill this finding. Reason: [explanation]. Move on." **DOWNGRADE:** "Q6 only shows technical possibility. Downgrade from High to Medium. Requires showing actual data exfil in PoC."
© Sachin Sharma · MIT · elementalsouls/Claude-BugHunter