SRC Bug Bounty Hunting
Overview
Security Response Centers (SRCs, 安全应急响应中心) run scoped, rules-based
bug bounty programs with payout tables, explicit out-of-scope lists, and
strict compliance red lines. Hunting them productively is a staged campaign —
intake, recon, enumeration, hunt, report — where the highest-value failures
are scope violations and evidence gaps, not missed vulnerabilities.
This skill encodes that campaign structure plus China-specific asset
fingerprinting (OA/中间件 identification, ICP-style asset mapping) and the
compliance discipline SRC programs enforce. It complements general
bug-hunting skills with the scope/payout discipline and time-boxed
prioritization that SRC platforms demand.
Source: cherry-picked and translated from zhaoxuya520/reverse-skill
(skills/pentest-tools/src-hunter, MIT license). Reference payload
libraries are not bundled; this skill is the standalone workflow.
When to Use
Trigger phrases:
- "hunt on this SRC program"
- "SRC bounty campaign"
- "scope-in/scope-out for a Chinese SRC"
- "time-boxed vulnerability hunt"
- "report a finding to an SRC"
- "China asset fingerprinting for bounty"
Use this skill when:
- You are working a scoped SRC/bug bounty target with published rules and
payout tables.
- You need a repeatable, evidence-disciplined campaign rather than ad hoc
probing.
- Out-of-scope boundaries and compliance red lines must be respected.
Prerequisites
- Scope authorization for every asset tested (see red lines below).
- A recon toolset: subdomain enumeration, HTTP probing, directory/parameter
fuzzing, and a vulnerability scanner for confirmation.
- The program's rules page: scope, out-of-scope, payout table, reporting
format.
Workflow
Phase 1: Intake
- Record the scoped target list, the out-of-scope list, the rules, and the
payout table in a campaign notes file.
- Note submission format and any "do not test" conditions (production
hours, rate limits, WAF behavior).
- Define the campaign time box and the target classes in scope.
Phase 2: Recon
- Subdomain discovery: passive (certificate transparency, DNS) then active
(brute force) enumeration; validate resolvable, live hosts.
- Perimeter mapping: CDN/WAF identification (knowledge-pass these — see red
lines), origin discovery, cloud asset inventory.
- File-browser and API discovery: exposed directories, JS bundle endpoints,
API documentation endpoints, and versioned API surfaces.
- China-specific fingerprinting: OA (office automation) systems,
中间件 (middleware: Weblogic, Tomcat, Nacos, Spring, etc.) version
identification — these are frequent SRC fast wins.
Phase 3: Enumeration
- Directory probing: dictionary + wordlist against live hosts; log status
code, size, and tech stack.
- Parameter dimension: enumerate parameters on interesting endpoints
(including hidden/undocumented params), record param types.
- Functional audit: walk each in-scope feature (auth, upload, search,
export, password reset) and note trust boundaries.
- Aim-tagging: tag each finding candidate with the vulnerability class it
most plausibly maps to — this drives Phase 4 priority.
Phase 4: Hunt
Prioritize by payout × exploitability using the standard class table:
| Priority |
Classes |
| High |
XSS (stored/reflected), RCE, SSRF, IDOR/BOLA, SQLi |
| Medium |
CSRF on sensitive actions, Path Traversal, File Upload, SSTI, XXE |
| Low |
Race conditions, HTTP Smuggling, OAuth/JWT/SAML misconfig, GraphQL abuse, Mobile API issues, LLM prompt-level flaws, DoS (scope-checked) |
For each candidate:
- Confirm the class with a minimal proof (no destructive payloads).
- Capture evidence: request/response pair, time, affected asset, impact
chain.
- Rule out false positives before moving on; a confirmed finding beats ten
unconfirmed leads.
Phase 5: Report
- Evidence discipline: every claim maps to a reproducible request/response
or trace — no screenshots of "trust me".
- Classification: state severity, vulnerability class, and business impact
(which data/function is affected).
- Relation to known issues: search the program's public duplicates before
submitting.
- Timeline: submission date, expected triage SLA, follow-up cadence.
Compliance Red Lines
- No production-damaging actions: no data destruction, no account takeover
beyond the minimal proof, no load-generating tests without authorization.
- No privacy exfiltration: never download or exfiltrate real user data;
use placeholders and minimal samples.
- No DoS: no volume/amplification testing.
- Authorization scope only: never test out-of-scope or third-party assets
reachable from an in-scope host.
- CDN/WAF knowledge-pass: identify them in recon and route around rather
than attacking the protection layer itself.
- Contact the SRC before testing anything unusual (origin discovery,
production-adjacent assets, third-party integrations).
Hands-On Example
Quick reachability probe of in-scope assets before active fingerprinting
(Phase 2 — Recon). With Python's httpx, installed via pip install httpx:
import httpx
for url in ["https://example.com"]:
r = httpx.get(url, timeout=10, headers={"User-Agent": "probe"})
print(url, r.status_code, len(r.content), r.headers.get("server"))
# https://example.com 200 559 cloudflare
Output above verified live (example.com 200, 559 bytes). A 200 from a
CDN-backed asset is a probe result, not a finding: log it, fingerprint the
origin — Kali's apt install nuclei installs nuclei 3.8+ — and only assets
that survive scope + compliance red-line checks get active testing.
Verification
Run this self-check before claiming completion:
When NOT to Use
- Unscoped testing or assets without authorization — stop and get scope.
- Pure infrastructure/networking programs (no web surface) — use network
pentest workflows.
- A program-agnostic hunt with no SRC rules or payout table —
bug-hunting
is the generic AI-orchestrated workflow.
- Long-running 24/7 scanning across many targets —
continuous-hunter;
SRC hunting is time-boxed by design.
- Long-horizon APT-style engagement — SRC hunting is time-boxed by design.
Anti-Rationalization Table
| Rationalization |
Reality |
| "It's in-scope-adjacent, so it's probably fine." |
Adjacent is out-of-scope. Test only listed assets; document the scope check per asset. |
| "I'll grab a real user record to prove impact." |
Privacy exfiltration is a red line and a program-ban offense. Use placeholders. |
| "One more payload pass on this low-priority class." |
Time-box discipline beats volume: spend the next pass on the payout-weighted table. |
| "The screenshot proves it." |
A screenshot without a request/response pair is not reproducible evidence; SRC triage will reject it. |
| "The WAF blocks me, so I'll attack the WAF." |
Knowledge-pass the protection layer; route around it or report the bypass properly. |
1---2name: src-hunter3description: 'Use when structured SRC bug bounty hunting for security response centers:4 intake scoped targets and payout rules, recon subdomains and perimeters, enumerate5 attack surface, hunt prioritized vulnerability classes with time-boxed discipline,6 and report with evidence discipline. Use when hunting on SRC platforms, running7 time-boxed bounty campaigns, or applying China-specific asset fingerprinting.'8domain: cybersecurity9subdomain: web-application-security10tags:11- bug-bounty12- src13- reconnaissance14- web-application-security15- pentesting16- vulnerability-hunting17version: '1.0'18author: oyi7719license: Apache-2.020nist_csf:21- DE.AE-0222- ID.RA-0123- RS.AN-0324category: cybersecurity25metadata:26 source: https://skillmd.com/skills/oyi77/src-hunter27---282930# SRC Bug Bounty Hunting3132## Overview3334Security Response Centers (SRCs, 安全应急响应中心) run scoped, rules-based35bug bounty programs with payout tables, explicit out-of-scope lists, and36strict compliance red lines. Hunting them productively is a staged campaign —37intake, recon, enumeration, hunt, report — where the highest-value failures38are scope violations and evidence gaps, not missed vulnerabilities.3940This skill encodes that campaign structure plus China-specific asset41fingerprinting (OA/中间件 identification, ICP-style asset mapping) and the42compliance discipline SRC programs enforce. It complements general43bug-hunting skills with the scope/payout discipline and time-boxed44prioritization that SRC platforms demand.4546Source: cherry-picked and translated from `zhaoxuya520/reverse-skill`47(`skills/pentest-tools/src-hunter`, MIT license). Reference payload48libraries are not bundled; this skill is the standalone workflow.4950## When to Use5152**Trigger phrases:**53- "hunt on this SRC program"54- "SRC bounty campaign"55- "scope-in/scope-out for a Chinese SRC"56- "time-boxed vulnerability hunt"57- "report a finding to an SRC"58- "China asset fingerprinting for bounty"5960Use this skill when:6162- You are working a scoped SRC/bug bounty target with published rules and63 payout tables.64- You need a repeatable, evidence-disciplined campaign rather than ad hoc65 probing.66- Out-of-scope boundaries and compliance red lines must be respected.6768## Prerequisites6970- Scope authorization for every asset tested (see red lines below).71- A recon toolset: subdomain enumeration, HTTP probing, directory/parameter72 fuzzing, and a vulnerability scanner for confirmation.73- The program's rules page: scope, out-of-scope, payout table, reporting74 format.7576## Workflow7778### Phase 1: Intake7980- Record the scoped target list, the out-of-scope list, the rules, and the81 payout table in a campaign notes file.82- Note submission format and any "do not test" conditions (production83 hours, rate limits, WAF behavior).84- Define the campaign time box and the target classes in scope.8586### Phase 2: Recon8788- Subdomain discovery: passive (certificate transparency, DNS) then active89 (brute force) enumeration; validate resolvable, live hosts.90- Perimeter mapping: CDN/WAF identification (knowledge-pass these — see red91 lines), origin discovery, cloud asset inventory.92- File-browser and API discovery: exposed directories, JS bundle endpoints,93 API documentation endpoints, and versioned API surfaces.94- China-specific fingerprinting: OA (office automation) systems,95 中间件 (middleware: Weblogic, Tomcat, Nacos, Spring, etc.) version96 identification — these are frequent SRC fast wins.9798### Phase 3: Enumeration99100- Directory probing: dictionary + wordlist against live hosts; log status101 code, size, and tech stack.102- Parameter dimension: enumerate parameters on interesting endpoints103 (including hidden/undocumented params), record param types.104- Functional audit: walk each in-scope feature (auth, upload, search,105 export, password reset) and note trust boundaries.106- Aim-tagging: tag each finding candidate with the vulnerability class it107 most plausibly maps to — this drives Phase 4 priority.108109### Phase 4: Hunt110111Prioritize by payout × exploitability using the standard class table:112113| Priority | Classes |114|---|---|115| High | XSS (stored/reflected), RCE, SSRF, IDOR/BOLA, SQLi |116| Medium | CSRF on sensitive actions, Path Traversal, File Upload, SSTI, XXE |117| Low | Race conditions, HTTP Smuggling, OAuth/JWT/SAML misconfig, GraphQL abuse, Mobile API issues, LLM prompt-level flaws, DoS (scope-checked) |118119For each candidate:120121- Confirm the class with a minimal proof (no destructive payloads).122- Capture evidence: request/response pair, time, affected asset, impact123 chain.124- Rule out false positives before moving on; a confirmed finding beats ten125 unconfirmed leads.126127### Phase 5: Report128129- Evidence discipline: every claim maps to a reproducible request/response130 or trace — no screenshots of "trust me".131- Classification: state severity, vulnerability class, and business impact132 (which data/function is affected).133- Relation to known issues: search the program's public duplicates before134 submitting.135- Timeline: submission date, expected triage SLA, follow-up cadence.136137## Compliance Red Lines138139- No production-damaging actions: no data destruction, no account takeover140 beyond the minimal proof, no load-generating tests without authorization.141- No privacy exfiltration: never download or exfiltrate real user data;142 use placeholders and minimal samples.143- No DoS: no volume/amplification testing.144- Authorization scope only: never test out-of-scope or third-party assets145 reachable from an in-scope host.146- CDN/WAF knowledge-pass: identify them in recon and route around rather147 than attacking the protection layer itself.148- Contact the SRC before testing anything unusual (origin discovery,149 production-adjacent assets, third-party integrations).150151## Hands-On Example152153Quick reachability probe of in-scope assets before active fingerprinting154(Phase 2 — Recon). With Python's httpx, installed via `pip install httpx`:155156```python157import httpx158for url in ["https://example.com"]:159 r = httpx.get(url, timeout=10, headers={"User-Agent": "probe"})160 print(url, r.status_code, len(r.content), r.headers.get("server"))161# https://example.com 200 559 cloudflare162```163164Output above verified live (example.com 200, 559 bytes). A 200 from a165CDN-backed asset is a probe result, not a finding: log it, fingerprint the166origin — Kali's `apt install nuclei` installs nuclei 3.8+ — and only assets167that survive scope + compliance red-line checks get active testing.168169## Verification170171Run this self-check before claiming completion:172173- [ ] Campaign notes contain scope, out-of-scope, rules, and payout table.174- [ ] Recon artifacts (subdomain list, live hosts, CDN/WAF map) exist and175 are reproducible.176- [ ] Every finding has a request/response evidence pair and an affected177 asset.178- [ ] Priority ordering followed payout × exploitability, not convenience.179- [ ] No red line was crossed; scope checks are documented per tested asset.180- [ ] Report includes classification, business impact, and duplicate check.181182## When NOT to Use183184- Unscoped testing or assets without authorization — stop and get scope.185- Pure infrastructure/networking programs (no web surface) — use network186 pentest workflows.187- A program-agnostic hunt with no SRC rules or payout table — `bug-hunting`188 is the generic AI-orchestrated workflow.189- Long-running 24/7 scanning across many targets — `continuous-hunter`;190 SRC hunting is time-boxed by design.191- Long-horizon APT-style engagement — SRC hunting is time-boxed by design.192193## Anti-Rationalization Table194195| Rationalization | Reality |196|---|---|197| "It's in-scope-adjacent, so it's probably fine." | Adjacent is out-of-scope. Test only listed assets; document the scope check per asset. |198| "I'll grab a real user record to prove impact." | Privacy exfiltration is a red line and a program-ban offense. Use placeholders. |199| "One more payload pass on this low-priority class." | Time-box discipline beats volume: spend the next pass on the payout-weighted table. |200| "The screenshot proves it." | A screenshot without a request/response pair is not reproducible evidence; SRC triage will reject it. |201| "The WAF blocks me, so I'll attack the WAF." | Knowledge-pass the protection layer; route around it or report the bypass properly. |