---

name: pentesting description: > Use this skill when the user asks about penetration testing, pentest, security assessment, vulnerability assessment, bug bounty, ethical hacking, or red team operations. version: "1.1.0" author: "j4flmao" license: "MIT" compatibility: claude-code: true cursor: true codex: true windsurf: true tags: [management, pentesting, phase-8]

Penetration Testing

Purpose

Produce structured penetration testing reports with methodology, severity-classified findings, evidence format, and remediation tracking.

Agent Protocol

Trigger

User request includes: pentest, penetration test, security test, bug bounty, vulnerability assessment, ethical hacking, red team, security assessment.

Input Context

• Scope (internal/external, web/mobile/api/infrastructure)
• Previous pentest reports (if any)
• Threat model and security requirements
• Compliance framework (PCI-DSS, SOC2, HIPAA)

Output Artifact

A markdown document containing:

• Pentest methodology (OSSTMM, OWASP, PTES)
• Test scope and exclusions
• Findings table with severity, CVSS, status
• Each finding: title, description, impact, reproduction steps, screenshot/evidence, recommendation
• Risk scoring methodology
• Remediation timeline with owner

Response Format

Produce the artifact directly. No preamble. No postamble. No explanations. No filler/hedging/transitions. Compress output — why use many token when few do trick.

---

Completion Criteria

• Methodology documented with references
• All findings classified by severity with CVSS scores
• Each finding includes: reproduction steps, evidence, remediation
• Executive summary for non-technical stakeholders
• Remediation tracking with SLA per severity

Max Response Length

4096 tokens

Workflow

Step 1: Define Scope

Document in-scope and out-of-scope targets, testing methodology (black/grey/white box), timeframe, and tools.

Step 2: Execute Tests

Follow OWASP/OSSTMM methodology, covering injection, auth, access control, XSS, SSRF, misconfiguration, crypto, and info leakage.

Step 3: Classify Findings

Score each finding using CVSS 3.1 and assign severity (Critical/High/Medium/Low/Info).

Step 4: Write Executive Summary

Summarize scope, overall risk rating, key findings count, and business impact for non-technical stakeholders.

Step 5: Document Each Finding

Each finding includes: title, severity, CVSS score, status, location, description, impact, reproduction steps, evidence, recommendation.

Step 6: Assign Remediation

Map each finding to an owner with a fix timeline per severity SLA.

Step 7: Threat Modeling

Perform threat modeling using STRIDE or PASTA before testing. Identify assets, threat agents, attack vectors, and security controls. Prioritize testing based on threat model findings. Map test cases to threats.

Step 8: Reconnaissance and Enumeration

Gather information about targets: DNS enumeration, subdomain discovery, technology fingerprinting, endpoint discovery, directory brute force, parameter discovery. Document all discovered surface area.

Step 9: Vulnerability Analysis

Correlate findings across test types. Eliminate false positives. Chain vulnerabilities to demonstrate realistic impact. Validate each finding manually. Prioritize chained attack paths.

Step 10: Reporting and Remediation Tracking

Generate structured report. Hold findings review with development team. Track remediation in issue tracker. Schedule retesting after fixes. Update risk register.

Framework / Methodologies

Pentest Methodology Comparison

Aspect	OWASP	PTES	OSSTMM	NIST SP 800-115
Focus	Web applications	Full penetration test	Security metrics	Technical assessment
Structure	12 testing categories	7 phases	5 channels (physical, human, wireless, telecom, data)	4 phases
Scoring	Risk rating methodology	CVSS + custom	STAR (Security Test Audit Report)	N/A
Best For	Web app pentesting	Comprehensive pentest	Audit/compliance	Government/enterprise
Maturity	Highly mature	Moderate	Very mature	Highly mature
Industry adoption	Most widely adopted	Common	Less common	Government standard

Decision Tree: Pentest Type Selection

What is being tested?
  ├── Web application
  │   ├── Internal app → Grey box, OWASP methodology
  │   └── External app → Black box + grey box, OWASP, DAST + manual
  ├── Mobile application
  │   ├── iOS → IPA analysis, runtime manipulation, network interception
  │   └── Android → APK analysis, rooted device testing, intent fuzzing
  ├── API / Microservices
  │   └── Grey box, OpenAPI-based testing, auth testing, injection, rate limiting
  ├── Cloud infrastructure
  │   ├── AWS → IAM analysis, S3 permissions, CloudTrail audit
  │   ├── GCP → IAM, GKE security, Cloud SQL audit
  │   └── Azure → RBAC, Key Vault, App Service, AKS security
  ├── Network infrastructure
  │   ├── Internal → Lateral movement, AD security, network segmentation
  │   └── External → Port scanning, service enumeration, exploit testing
  └── Physical / Social engineering
      └── OSSTMM methodology, physical access testing, phishing simulation

Testing Types Comparison

Type	Reconnaissance	Access	Knowledge	Accuracy	Coverage	Cost
Black box	Full	None	None	Medium	Low	High
Grey box	Targeted	Some	Architecture	High	Medium	Medium
White box	Technical docs	Full access	Full code	Very high	High	Low
DAST	None	None	None	High	Medium	Low
SAST	Full source	Full code	Full	Medium	High	Low-medium
Bug bounty	None	None	None	High	Variable	Pay per finding

Common Pitfalls

Pitfall 1: Testing Without a Clear Scope

Starting a pentest without clearly documented scope leads to disputes about what was included, missed targets, and wasted effort. Always define in-scope and out-of-scope in writing. Include URL/IP lists and explicit exclusions.

Pitfall 2: Relying Only on Automated Scanners

Automated scanners produce high false-positive rates and miss logic flaws, business logic vulnerabilities, and chained attacks. DAST tools only catch known patterns. Manual testing is essential for high-severity findings.

Pitfall 3: Not Validating Findings

Reporting scanner output without manual validation produces reports full of false positives that teams learn to ignore. Manually validate every finding before including it in the report. Document validation steps in the finding.

Pitfall 4: Incomplete Evidence Documentation

Findings without clear reproduction steps, request/response pairs, or screenshots cannot be reproduced or fixed. Document evidence as you go — recreating evidence later is time-consuming and error-prone.

Pitfall 5: CVSS Without Vector String

Reporting a CVSS score without the vector string makes the score meaningless — it cannot be verified or recalculated if the environment changes. Always include the full vector string.

Pitfall 6: Not Testing Authentication and Authorization Thoroughly

Many testers check basic login/logout but miss IDOR, privilege escalation, session fixation, and JWT vulnerabilities. These are often the highest-impact findings. Dedicate significant testing time to auth and access control.

Pitfall 7: Ignoring Business Logic

Automated scanners don't test business logic — multi-step workflows, approval chains, financial calculations, and user role interactions. Manual business logic testing frequently finds critical vulnerabilities.

Pitfall 8: Reporting Without Remediation Guidance

Finding vulnerabilities without actionable remediation guidance frustrates development teams and delays fixes. Every finding must include specific, actionable recommendation code snippets where possible.

Pitfall 9: Not Sanitizing Evidence

Including live credentials, real customer data, or internal IP addresses in reports creates a secondary security incident. Sanitize all evidence before including it in the report.

Pitfall 10: Missing Retesting After Fixes

Remediation verification is as important as initial testing. Without retesting, you cannot confirm that fixes are effective or that they haven't introduced new issues. Schedule retesting as part of the engagement.

Best Practices

• Define scope in writing before testing begins: Document in-scope targets, testing methods, time window, and emergency contacts. Get written authorization.
• Use a consistent methodology: Follow OWASP or PTES formally. Document which tests were performed and which were not.
• Score every finding with CVSS 3.1 including vector string: Enables verification, recalculation, and consistent severity communication.
• Validate every finding manually: Automated findings are leads, not confirmed vulnerabilities. Manual validation separates real issues from false positives.
• Document reproduction steps in detail: A developer should be able to reproduce and verify the finding from your documentation alone.
• Chain vulnerabilities to demonstrate real impact: A low-severity info leak combined with another finding may create a critical risk.
• Write executive summaries for non-technical stakeholders: Business impact, risk level, and remediation priorities — no technical jargon.
• Sanitize all evidence: Remove credentials, tokens, internal IPs, and customer data before including in reports.
• Schedule retesting as part of the engagement: Initial fix is not enough — verify effectiveness and check for regression.
• Track findings in the project's issue tracker: Integrate with development workflow for assignment, prioritization, and closure tracking.
• Establish remediation SLAs per severity: Critical in 24 hours, High in 7 days, Medium in 30 days, Low in 90 days.
• Maintain a findings library: Reusable finding descriptions save time and ensure consistency across engagements.
• Update methodology regularly: The threat landscape evolves. Review and update test cases quarterly.

Templates & Tools

Pentest Scope Definition

engagement:
  title: External Web Application Pentest
  client: Acme Corp
  date: 2026-04-15 to 2026-04-26
  methodology: OWASP Web Application Testing Guide v4.2
  testing_type: Grey box
  in_scope:
    targets:
      - https://app.acme.com
      - https://api.acme.com/v2
      - https://admin.acme.com
    testing_allowed:
      - Authentication and session testing
      - Injection testing
      - Business logic testing
      - Access control testing
  out_of_scope:
    - Third-party services (Auth0, Stripe, AWS)
    - Denial of Service attacks
    - Physical security
    - Social engineering
    - Employee accounts (testing accounts provided)
  rules_of_engagement:
    - Testing hours: 08:00 - 18:00 local time only
    - No destructive testing without approval
    - Emergency contact: security@acme.com / +1-555-0123
    - Report critical findings immediately, not in final report
    - Data handling: all data destroyed 30 days after report delivery

Finding Template

finding:
  id: PENTEST-2026-001
  title: SQL Injection in Order Search
  severity: Critical
  cvss: "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H"
  cvss_score: 9.8
  status: Open
  location: "POST /api/v2/orders/search (order-service)"
  discovered: 2026-04-16
  description: |
    The `q` parameter in the order search endpoint is directly concatenated
    into SQL queries without parameterization or sanitization.
  impact: |
    An unauthenticated attacker can extract, modify, or delete all database
    records, including user credentials, payment information, and order data.
  reproduction:
    - Step 1: Intercept request to POST /api/v2/orders/search
    - Step 2: Set payload: q=1' OR '1'='1
    - Step 3: Observe that all orders are returned instead of user's orders
    - Step 4: Extract data using UNION-based injection
  evidence:
    request: "POST /api/v2/orders/search HTTP/1.1\nContent-Type: application/json\n\n{\"q\":\"1' OR '1'='1\"}"
    response: "HTTP/1.1 200 OK\n... (150 orders returned, including orders from other users)"
  recommendation: |
    Replace string concatenation with parameterized queries:
    ```python
    cursor.execute("SELECT * FROM orders WHERE user_id = %s AND search LIKE %s",
                   (user_id, f"%{search_term}%"))
    ```
  remediation:
    owner: team-orders
    due_date: 2026-04-17
    verification: Re-test after deployment to staging
  references:
    - https://owasp.org/www-community/attacks/SQL_Injection
    - https://cheatsheetseries.owasp.org/cheatsheets/SQL_Injection_Prevention_Cheat_Sheet.html

Tool Selection Matrix

Tool Category	Recommended Tools	Use Case	Cost
Scanner (web)	Burp Suite Professional, OWASP ZAP	Web app scanning	$ / Free
Scanner (network)	Nessus, OpenVAS	Network vulnerabilities	$ / Free
Scanner (cloud)	ScoutSuite, Prowler, CloudSploit	Cloud config review	Free
Reconnaissance	Amass, Sublist3r, ffuf, gau, httpx	Subdomain and endpoint discovery	Free
Exploitation	Metasploit, Cobalt Strike	Post-exploitation	Free / $$
Mobile	MobSF, Frida, objection, jadx	Mobile app analysis	Free
Wireless	Aircrack-ng, Kismet, Reaver	Wireless security	Free
Password	Hashcat, John the Ripper, Hydra	Password cracking	Free
OSINT	Maltego, theHarvester, Shodan	Intelligence gathering	Free / $
DAST	Burp Scanner, Acunetix, Netsparker	Automated web scanning	$ / $$
SAST	Semgrep, CodeQL, SonarQube	Source code analysis	Free / $
API	Postman, Insomnia, Burp	API security testing	Free / $

Case Studies

Case Study 1: E-Commerce SQL Injection

During a grey-box pentest of an e-commerce platform, the tester found a SQL injection in the order search endpoint. The automated scanner flagged the parameter but the test team validated it manually by extracting database version and table structure. Chaining this with an information disclosure finding (debug endpoint exposed database schema) demonstrated complete database compromise. Remediation: parameterized queries implemented within 4 hours of the finding being reported.

Case Study 2: Cloud Misconfiguration Leading to Data Breach

An infrastructure pentest of an AWS environment using ScoutSuite found 47 S3 buckets with public read access. Manual validation confirmed that 12 buckets contained PII (customer names, emails, purchase history). The finding was escalated to CISO within 24 hours. Remediation: S3 bucket policies updated, CloudTrail enhanced, and automated compliance scanning implemented.

Case Study 3: Business Logic Flaw in Payment Processing

During an API pentest, manual business logic testing revealed that a payment API did not validate that the requesting user owned the order they were paying for. An attacker could change another user's order amount from $100 to $0.01 and pay the reduced amount. No automated scanner would catch this. Remediation: server-side ownership validation was added.

Case Study 4: JWT Authentication Bypass

A mobile app pentest using Frida and runtime manipulation discovered that the app accepted unsigned JWTs with arbitrary user IDs. The finding was validated by creating a JWT with {"sub": "admin", "role": "admin"} and no signature, which the server accepted. Root cause: the server verification library was misconfigured in "debug mode". Remediation: JWT signature verification properly configured, production configs reviewed.

Rules

• Every finding must include a CVSS 3.1 score with the vector string
• Reproduction steps must be detailed enough for another tester to replicate
• Evidence must be sanitized (no live credentials, customer data, or internal IPs)
• False positives must be documented with reasoning, not simply removed
• Retesting after fix must use the same methodology as the original test
• Remediation SLA starts from the report delivery date, not the finding date
• Critical findings require CISO notification within 24 hours of discovery
• Scope must be documented in writing before testing starts
• Automated findings must be manually validated before inclusion
• Business logic must be tested manually, not just via automated scanning
• Executive summary must be suitable for non-technical audience
• All evidence must be collected during testing, not reconstructed after
• Test methodology must be documented with references
• Remediation guidance must be actionable and specific

Report Structure

1. Executive Summary

• Scope overview
• Overall risk rating (Critical / High / Medium / Low)
• Key findings summary (count per severity)
• Business impact statement

2. Scope

• In-scope: URLs, IP ranges, API endpoints, mobile apps, source code
• Out-of-scope: explicitly listed
• Testing methodology (black/grey/white box)
• Timeframe and tools used

3. Risk Scoring

Severity	CVSS 3.1	Impact	Exploitability	Count
Critical	9.0-10.0	Complete compromise	Easy, public exploit	#
High	7.0-8.9	Significant data/access	Moderate	#
Medium	4.0-6.9	Limited impact	Complex	#
Low	0.1-3.9	Minimal	Very complex	#
Info	0.0	No direct risk	N/A	#

4. Finding Format

Each finding MUST contain:

[CRITICAL] SQL Injection in Order Search

ID: PENTEST-2026-001 Severity: Critical (CVSS 9.8) Status: Open / Fixed / Accepted / Mitigated Location: POST /api/v2/orders/search (order-service)

Description

The q parameter is directly concatenated into SQL query without parameterization.

Impact

An attacker can extract, modify, or delete all database records.

Reproduction

1. Intercept request to POST /api/v2/orders/search
2. Set payload: q=1' OR '1'='1
3. Observe that all orders are returned

Evidence

Request: POST /api/orders/search Payload: q=1' OR '1'='1 Response: 200 OK (150 orders returned, including orders from other users)

Recommendation

Replace string concatenation with parameterized queries.

Remediation

• Owner: team-orders
• Due: 2026-05-16
• Verification: Re-test after deployment

Remediation SLA

Severity	Fix Timeline	Verification	Escalation
Critical	24 hours	Re-test required	CISO
High	7 days	Re-test required	Security lead
Medium	30 days	Code review	Engineering manager
Low	90 days	Code review	Team lead

Common Findings Catalog

Category	Finding	Typical Severity
Injection	SQL, NoSQL, Command, LDAP injection	Critical/High
Auth	Broken authentication, session fixation, weak password policy	Critical/High
Access Control	IDOR, privilege escalation, missing function-level access control	High
XSS	Stored, Reflected, DOM-based	Medium/High
SSRF	Server-side request forgery	High
Misconfiguration	Default credentials, debug endpoints, CORS misconfig	Medium
Crypto	Weak TLS, hardcoded keys, weak hashing	Medium
Info Leakage	Stack traces, error messages, directory listing	Low/Info

References

• references/findings-library.md — Pentest Findings Library
• references/pentest-methodology.md — Pentest Methodology Reference
• references/pentesting-advanced.md — Pentesting Advanced Topics
• references/pentesting-fundamentals.md — Pentesting Fundamentals
• references/pentesting-methodology.md — Pentesting Methodology
• references/report-template.md — Pentest Report Template
• references/pentesting-methodology-guide.md — Comprehensive pentesting methodology guide
• references/pentesting-tools-automation.md — Pentesting tools and automation reference

Handoff

Hand off to management/security/SKILL.md for vulnerability remediation workflow. Hand off to management/alerting/SKILL.md for detection rule creation.

Architecture Decision Trees

Testing Approach

Decision Point	Option A	Option B	Decision Criteria
Knowledge level	Black box (realistic external)	White box (thorough coverage)	Budget, compliance requirements
Methodology	OWASP WSTG (web-focused)	PTES (comprehensive)	Application type, org maturity
Tool selection	Automated scanners (efficient)	Manual testing (deep findings)	Time constraints, finding quality needs

Scope Definition

• External-facing apps → Full external pentest with auth bypass attempts
• Internal apps → Internal network pentest with privilege escalation
• Mobile apps → API + device-side testing with traffic interception
• APIs → Authorization + injection + rate limiting tests

Implementation Patterns

Pentest Report Template

`markdown

Pentest Report: {target}

Executive Summary

• Scope: {systems tested}
• Duration: {dates}
• Finding count: {critical}/{high}/{medium}/{low}

Critical Findings

{finding title}

• CVE/CWE: {reference}
• CVSS: {score}
• Impact: {business risk}
• Reproduction: {steps}
• Evidence: {screenshot/log}
• Recommendation: {fix}

Risk Summary

Risk Level	Count	Examples
Critical	{n}	{titles}
High	{n}	{titles}
Medium	{n}	{titles}
Low	{n}	{titles}
`

Vulnerability Reproduction Steps

`markdown

SQL Injection in /api/users endpoint

1. Login as standard user
2. Intercept request to GET /api/users?id=1
3. Modify id parameter: 1 UNION SELECT username,password FROM admin_users
4. Observe admin credentials in response

Remediation

• Use parameterized queries
• Implement WAF rule to block SQLi patterns

Production Considerations

Rules of Engagement

• Authorization: Obtain signed authorization letter before testing. Define scope, methods, and exclusions explicitly.
• Communication: Designate emergency contact for immediate critical findings. Define escalation path for production impact.
• Rollback plan: Have system rollback procedures ready. Test during maintenance windows when possible.

Quality Assurance

• Validation: Verify each finding manually to eliminate false positives. Attempt exploitation to confirm impact.
• Peer review: Have findings reviewed by second pentester. Validate remediation steps are practical and complete.
• Retesting: Schedule retest after remediation within agreed timeframe. Test only fixed scope, not full re-pentest.

Anti-Patterns

Anti-Pattern	Symptom	Solution
Scanner-only testing	50+ false positives, no real findings	Always manually verify scanner findings
No retest window	Findings never get validated	Contractually require retest within 30 days
Scope creep	Missed critical areas due to unfocused testing	Strictly define scope and stick to it
Findings without business context	Management ignores report	Add business impact, regulatory risk, remediation cost
No evidence	Cannot reproduce findings	Screenshots, request/response pairs, proof of concept

Performance Optimization

Efficiency Techniques

• Automated reconnaissance: Use subdomain enumeration, port scanning, tech fingerprinting upfront. Prioritize attack surface.
• Tool chaining: Automate scanner output to feed manual testing. Use Burp Collaborator for out-of-band detection.
• Templated reports: Auto-generate report skeleton from tooling output. Manually enrich critical findings only.

Time Management

• Day 1-2: Reconnaissance and automated scanning. Gather intelligence on target.
• Day 3-4: Deep manual testing of critical functionality. Exploit automation findings.
• Day 5: Report writing, evidence collection, retest verification. Deliverable preparation.

Security Considerations

Operational Security

• Testing infrastructure: Use isolated VPS for scanning. Never conduct tests from corporate network without authorization.
• Data handling: Encrypt findings in transit and at rest. Never exfiltrate real customer PII — use synthetic test data.
• Tool security: Keep testing tools updated. Use reputable sources for tool downloads.

Confidentiality

• Report distribution: Share reports only with authorized stakeholders. Use DRM-protected PDF for distribution.
• Finding disclosure: Follow responsible disclosure timelines. Notify vendor before public disclosure of zero-days.
• Non-disclosure: Sign NDA before receiving target documentation. Maintain confidentiality of system details.