Recon Security

Recon Security

This skill guides an agent through an authorized external pentest workflow: recon, normalization, active discovery, web and infrastructure checks, validation, scoped exploitation (when RoE allows), and reporting. Use only free/open-source tools unless the user explicitly opts into commercial services outside this skill.

Do not assume permission. Gate every active, invasive, or state-changing step on Pass 0. This skill is model-guided only: propose commands and workflows; the user or agent runs them when scope and mode allow. No bundled scripts ship with this skill.

Engagement lifecycle

Pass 0 Scope/RoE → Pass 1 Passive → Pass 2 Normalize → Pass 3 Active
    → Pass 4 Web + infra → Pass 5 Triage → Pass 6 Validation
    → Pass 7 Scoped exploitation (if approved) → Pass 8 Report

Mental model

• Recon maps what is exposed.
• Validation proves what matters with minimal reproducible evidence.
• Exploitation demonstrates impact only within written RoE — not unrestricted attack.

Prefer conservative, reproducible checks. One confirmed finding beats dozens of scanner lines.

Pass 0: scope and authorization

Establish before any active work:

• In-scope domains, subdomains, IPs, ASNs, apps, and environments (prod vs staging).
• Out-of-scope assets and forbidden techniques (brute force, DoS, data dump, lateral movement).
• Written authorization or explicit ownership.
• Scan intensity: passive only, light active, standard, or deep.
• Whether exploitation and credential testing are allowed.
• Evidence directory (PROJECT_DIR, default ~/Projects/pentest-engagement). See references/environment-setup.md.
• Deliverable: command plan only, executed tests, or full report.

If authorization is unclear, stop at passive planning and ask.

Pass 1: passive recon

Build inventory from public sources only:

• DNS, WHOIS/RDAP, RIPEstat/BGPView, certificate transparency (crt.sh, subfinder, amass -passive, assetfinder).
• Historical URLs (gau, waybackurls, Common Crawl, public urlscan.io).
• Search dorks and public code references (no secret copying).

Use example commands from references/environment-setup.md and references/tools.md when the user wants executable steps.

Pass 2: normalize targets

Produce working lists under targets/:

• domains.txt, in_scope_domains.txt, resolved_hosts.txt, web_targets.txt, ips.txt
• needs-scope-confirmation.txt for uncertain assets

Deduplicate; drop unrelated CT names and out-of-scope SaaS unless approved.

Pass 3: active recon

Requires Pass 0 approval. Rate-limit all probes.

• dnsx, httpx, nmap (top ports first), optional naabu / rate-limited masscan
• wafw00f, testssl.sh or sslyze, nuclei (open templates; triage as leads)

Propose rate-limited active commands; save raw output under evidence/active/.

Pass 4: web app and infrastructure checks

Web applications

• Fingerprint headers, cookies, security headers, technologies.
• Misconfigurations: .git, .env, backups, phpinfo, directory listing, robots.txt.
• Content/parameter discovery: ffuf, feroxbuster, arjun, katana, hakrawler.
• Light automated probes; sqlmap detection-only by default; dalfox for reflected XSS leads.
• Manual proxy review: OWASP ZAP or Burp Suite Community.

Walk through fingerprinting, misconfiguration checks, and light probes for each priority URL.

Infrastructure (when in scope)

Telecom, storage, and file services often appear on external pentests:

• SIP/VoIP: UDP/TCP 5060/5061, nmap --script=sip-methods, OPTIONS probes. No call setup unless authorized.
• NAS/file exposure: ports 445, 139, 548, 873, 2049, 5000/5001, 8080; smbclient -N -L, showmount -e. Document share permissions, not customer file contents.

Save SIP/NAS results under evidence/infra/.

Pass 5: triage and evidence

Classify every item:

Build a prioritized queue for validation. Store under evidence/triage/. Never paste secrets or bulk PII into reports.

Pass 6: validation

Turn leads into confirmed findings. Read references/validation.md.

• Import web_targets.txt into Burp or ZAP; map auth and roles.
• Manually confirm nuclei/ffuf/sqlmap signals.
• Two-account testing for IDOR/BOLA; pair with authz-security when code is available.
• Infrastructure: focused port/service re-checks; SIP/NAS proof without data theft.

PoC bar: numbered steps, request/response or screenshot, impact, fix.

Pass 7: scoped exploitation

Only when Pass 0 explicitly allows exploitation. Read references/exploitation-roe.md.

• Minimum proof of impact (one row, one harmless upload, one auth bypass with test accounts).
• No --dump, persistence, lateral movement, or destructive actions unless contract permits.
• Remove test artifacts when cleanup is required.
• Stop and escalate if scope, production risk, or legal boundaries are unclear.

Pass 8: reporting

Use references/report-template.md. Include:

• Executive summary and scope
• Methodology by phase (passive, active, validation, exploitation if run)
• Findings by severity with reproduction and remediation
• Leads and limitations
• Remediation roadmap (immediate / short / long term)

Severity scale

• P0: Sensitive data exposure, unauthenticated admin/control, confirmed critical exploit path.
• P1: High-impact issue with limited preconditions; confirmed injection or authz break without mass extraction.
• P2: Medium exposure or hardening gap without confirmed exploit chain.
• P3: Informational, hygiene, or scan limitations.

Output format

[P1] exposed-admin-panel on https://admin.example.com
  Evidence: httpx + manual browser review; headers in evidence/webapp/.
  Impact: Public admin surface increases credential and exploit risk.
  Fix: Restrict by VPN/IdP, enforce MFA, monitor access.

Reference files

• references/tools.md — approved tools and commercial exclusions
• references/checklist.md — full engagement checklist
• references/environment-setup.md — macOS setup and directory layout
• references/validation.md — manual validation and PoC bar
• references/exploitation-roe.md — allowed/prohibited exploitation boundaries
• references/report-template.md — final deliverable structure

What this skill won't do

• Require Shodan, Censys, DeHashed, IntelX, or Burp Pro.
• Treat scanner output as confirmed without Pass 6.
• Run exploitation, dumping, persistence, or lateral movement without explicit RoE approval.
• Bypass authorization or test out-of-scope assets.