A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity + infrastructure attack matrices.
Python ai-securityanthropicapplication-securitybug-bountybugbountybugcrowdclaudeclaude-codeclaude-skillsethical-hackinghackeroneoffensive-securitypentestingred-teamsecurity-toolsweb-security
feat(skills): 11 disclosed-report technique gaps across 8 hunt skills (#73)
Sachin Sharma7c7210ee5ef5
diff --git a/skills/hunt-ato/SKILL.md b/skills/hunt-ato/SKILL.mdindex ad5a518..a1ffc0b 100644--- a/skills/hunt-ato/SKILL.md+++ b/skills/hunt-ato/SKILL.md@@ -114,6 +114,12 @@ dig +short staging.target.com # CNAME -> nonexistent-app.herokuapp.com (NXDO```**Confirmation = OOB:** the auth `code` (or implicit `access_token`) must actually arrive at the host you claimed — log it server-side and exchange it for B's token. A redirect_uri that merely *reflects* an off-origin value but bounces the code through a server-pinned exchange is not exploitable. Decode any error body as JSON, not substring — `AADSTS50076` / claims-challenge responses contain a literal `access_token` substring inside the claims field that is NOT a usable token.+### Path 10: Pre-Account / Unverified-Account Takeover+The victim's account exists in a pending/unverified state (invite not accepted, signup email never confirmed, or no email on file). Claim it first: register or attach your controlled email to that identity, then let the victim complete SSO/OTP login and inherit the attacker-seeded state — or trigger a reset once your email is attached. Test invite/pending flows and any endpoint that writes an email to an account lacking a confirmed one; confirm the victim, after signing in, lands in the account you seeded. High impact, no token interaction. Disclosed: reports/1679734, reports/394329.++### Path 11: Passwordless / Magic-Link Abuse by Identifier+A passwordless-signup / magic-login / OTP-login endpoint that sets or resets a password (or issues a session) given only an enumerable identifier (phone/email), with no token step. Probe `*passwordless*`, `*magic*`, `*/otp-login`, `*/signup`: does supplying only a phone/email set the password or return a session for that identity? Enumerate identifiers to prove arbitrary-account reach; validate by taking test account B via its number alone. Disclosed: reports/143717.+### ATO Severity Gate- **Critical** — zero/low victim interaction: Host-header reset poisoning, JWT forgery to victim endpoint, lax-redirect_uri auth-code theft, IDOR-driven email change → reset.- **High** — one email click OR a pre-existing session/cookie required (Referer leak, no-step-up password change behind cookie theft).diff --git a/skills/hunt-auth-bypass/SKILL.md b/skills/hunt-auth-bypass/SKILL.mdindex 159fa70..4d4d4e9 100644--- a/skills/hunt-auth-bypass/SKILL.md+++ b/skills/hunt-auth-bypass/SKILL.md@@ -20,6 +20,9 @@ Auth bypass is consistently one of the highest-paying vulnerability classes in b---+### SSH certificate-authority trust forgery (Git-hosting / enterprise platforms)+When an org or instance registers an SSH certificate authority, cert `principals` may not be bound server-side to the requesting identity — a member can mint a cert asserting another user's principal and authenticate as them (e.g. modify another user's resource given only its URL). Add SSH-CA trust to the auth-surface list on Git-hosting targets; niche/platform-specific. Disclosed: reports/1901040.+## Attack Surface Signals**URL patterns to hunt:**diff --git a/skills/hunt-brute-force/SKILL.md b/skills/hunt-brute-force/SKILL.mdindex ad406d4..9b18273 100644--- a/skills/hunt-brute-force/SKILL.md+++ b/skills/hunt-brute-force/SKILL.md@@ -163,6 +163,9 @@ curl -s -X POST "https://$TARGET/forgot-password" -d "email=$INVALID_USER" | grecurl -s -X POST "https://$TARGET/api/register" -d "email=$VALID_USER" | grep -i "exist\|taken\|already"```+### Phase 3b — Unthrottled registration (mass account creation)+An unrate-limited signup endpoint is an abuse finding on its own, not just an enumeration oracle: burst `*/signup`/`*/register` (rotate email + source per request) and confirm N accounts are actually created with no 429/CAPTCHA/lockout → automated mass-account creation (spam, promo/referral abuse, resource exhaustion). Low–Medium standalone; higher when it chains to a paid/limited resource. Disclosed: reports/2915502.+### Phase 4 — IP / Source Rotation Bypass```bash# Per-IP limits are bypassable when the app trusts a client-controlled source header.diff --git a/skills/hunt-cache-poison/SKILL.md b/skills/hunt-cache-poison/SKILL.mdindex 180f4db..374659c 100644--- a/skills/hunt-cache-poison/SKILL.md+++ b/skills/hunt-cache-poison/SKILL.md@@ -206,6 +206,9 @@ True-Client-IP---+### Origin header → ACAO cache poisoning+Add `Origin` to the unkeyed-header set. Test whether it is reflected into `Access-Control-Allow-Origin` and then cached: `curl -H "Origin: evil.example" URL -I` on two consecutive hits. A cached attacker/`*` ACAO breaks CORS for legit users (cache-DoS); a cached permissive ACAO enables cross-user data reads. Disclosed: reports/591302.+## Common Root Causes1. **CDN misconfiguration — caching based on URL path only.** Engineers configure cache rules like "cache everything matching `*.js`" without realizing the path can be appended to dynamic routes. The origin server ignores the extra path segments, but the CDN uses them as cache keys.diff --git a/skills/hunt-cloud-misconfig/SKILL.md b/skills/hunt-cloud-misconfig/SKILL.mdindex 2d4e75d..9c9b67b 100644--- a/skills/hunt-cloud-misconfig/SKILL.md+++ b/skills/hunt-cloud-misconfig/SKILL.md@@ -37,6 +37,12 @@ http://169.254.169.254/latest/meta-data/iam/security-credentials/ROLE-NAME # ke---+### Jenkins SSO-restriction failure → Script Console RCE+An exposed Jenkins that authenticates ANY personal Google/OAuth account (missing org-domain restriction) is full compromise: log in, then `/script` (Groovy console) = RCE, and `/credentials` + `/scriptText` = stored API tokens and source disclosure. Disclosed: reports/231460.++### Exposed message broker (RabbitMQ / AMQP) with default creds+Probe RabbitMQ management (`http://TARGET:15672/`, API `:15672/api/overview`) and AMQP `:5672`; try default `guest:guest` (works off-localhost when `loopback_users` is misconfigured) → queue/message read, vhost enumeration, management-plugin admin. Disclosed: reports/753602.+## Local-verification toolchainFor testing cloud-misconfig findings against a local AWS sim before/instead of hitting real cloud:diff --git a/skills/hunt-csrf/SKILL.md b/skills/hunt-csrf/SKILL.mdindex b6954ec..cf98d14 100644--- a/skills/hunt-csrf/SKILL.md+++ b/skills/hunt-csrf/SKILL.md@@ -267,6 +267,9 @@ curl -s https://monitoring.target.com/api/health | jq '.version'---+### Mobile-app CSRF via custom-scheme deeplink+CSRF is not browser-only. For targets with a mobile app, enumerate exported deeplink / custom-scheme handlers (`scheme://action` from the Android manifest / iOS `Info.plist`) and test whether opening one performs a state-changing action (follow, link, purchase) using the app's ambient session with no CSRF token — deliverable to a victim by a plain link or a QR code. Disclosed: reports/583987, reports/805073.+## Gate 0 Validation1. **What can the attacker DO right now?** — The attacker must be able to trigger a specific state-changing action (account linking, email change, data deletion, social association) on behalf of the victim without any interaction beyond visiting a URL or page.diff --git a/skills/hunt-dom/SKILL.md b/skills/hunt-dom/SKILL.mdindex 423de1e..9c43604 100644--- a/skills/hunt-dom/SKILL.md+++ b/skills/hunt-dom/SKILL.md@@ -134,6 +134,9 @@ addEventListener('message', e => {---+### PostMessage handler race — win the message before origin is set+Beyond static origin/sender checks: some handlers trust the FIRST message, or read `event.origin` while it is still `""`/`null` during frame load. Race it — flood `postMessage` in a tight loop during iframe/popup load, or fire from `about:blank` before navigation, so your message lands before the legit sender or before strict origin setup runs. Test any handler that assumes "first message = trusted". Disclosed: reports/381356.+## Phase 3 — Service Worker Abuse**Hard rule (corrects a common mistake):** a SW script URL **must be same-origin** as the page calling `register()`. A cross-origin script URL (`https://evil.com/sw.js`) throws `SecurityError` — there is **no header that enables cross-origin SW *script* registration**. `Service-Worker-Allowed` only widens the **scope** a same-origin script may control, not where the script may live.diff --git a/skills/hunt-lfi/SKILL.md b/skills/hunt-lfi/SKILL.mdindex 3a29bec..9203a98 100644--- a/skills/hunt-lfi/SKILL.md+++ b/skills/hunt-lfi/SKILL.md@@ -169,6 +169,11 @@ curl -s "https://$TARGET/" -H "User-Agent: <?php system(\$_GET['c']); ?>"?file=zip:///var/www/uploads/a.zip%23path/inside.txt```+### Phase 8b — Archive-extraction & image-processing sinks (no `?file=` needed)+Two LFI/traversal surfaces that are not query-parameter file reads:+- **Zip Slip / tar-symlink escape** — a server that untars/unzips a user-supplied archive follows `../../` entry names or embedded symlinks (tar preserves them) to read/write outside the extraction dir. Craft an archive entry named `../../../etc/cron.d/x`, or a symlink pointing at `/etc/passwd`; confirm by writing a unique canary outside the upload dir. Disclosed: reports/1439593, reports/733072, reports/822262.+- **ImageMagick coder read-sink** — image convert/thumbnail/avatar features read local files via MSL/MVG coders even with no file parameter: filename/label `label:@/etc/passwd`, or an MSL payload (`<image xlink:href="msl:/etc/passwd">`) plus `msl:`/`ephemeral:` pseudo-protocols (ImageTragick CVE-2016-3714 family). Disclosed: reports/1858574. Pairs with `hunt-file-upload`.+### Phase 9 — Automation (then manual-confirm everything)```bashffuf -u "https://$TARGET/page.php?file=FUZZ" -w ~/wordlists/lfi.txt -mc all -fr "not found"