Agent Skills: Pentest Skill

Pentest Skill

[!IMPORTANT] PTES-aligned adversarial security audit for backend, frontend, and mobile applications. Produces a CVSS-scored Hacker Report with verified PoCs and phased remediation.

Optional args: slug=<feature>, ticket=<id/url>, mode=interactive|autonomous|channel, channel=<id>, auto_continue=true|false, profile=business|hybrid|technical.

Instructions

When the user asks to perform this workflow, execute the following steps:

🕵️‍♂️ Penetration Test (PTES-Aligned)

Goal: Execute a red-team assessment across backend, frontend, and mobile targets with verified PoCs and audit-grade evidence.

Steps

1. Scope and authorization:

   • Determine mode (whitebox, greybox, blackbox) and targets.
   • Verify explicit, current authorization, target exclusions, permitted techniques, time window and stop contact. Require demonstrated runtime controls for live actions; unsupported controls block execution, not offline planning.

2. Intel and threat model:

   • Identify stack, endpoints, parameters, local storage, schemes, and secrets.
   • Use common-security-audit, common-dast-tooling, and architecture docs when available.
   • Map trust boundaries, secrets, external integrations, and privileged workflows before exploit attempts.

3. Vulnerability analysis:

   • Run scans and delegate: SAST/SCA to specialist-aspm-correlator, dynamic/logic to specialist-logic-hacker, binary/mobile to specialist-mobile-reverser.
   • Rank targets by exposure × sensitivity × auth_coverage.

4. Exploit verification:

   • Confirm findings only with reproducible evidence within scope; never force unsafe exploitation.
   • Optional lab handoff: When verifying an exploit hypothesis against an isolated synthetic lab target, hand off to cyber-lab-validation with action-specific authorization and verified host isolation controls.
   • Prohibitions: Never target production/public assets, use real credentials, or exfiltrate real data.
   • If active verification is blocked, refused, or unsupported, record step blocked or finding not-tested; continue independent threat modeling, code review, and remediation guidance without retrying with disguised prompts.
   • Preserve suspected, blocked, not-tested and false-positive dispositions separately from confirmed; record missing evidence, preconditions and blast radius.
   • Use synthetic identities/data; no real sensitive-data extraction. Stop on scope drift or impact; obtain approval before restarting.

5. Reporting:

   • Write artifacts/security-review.md with assumptions, source provenance, review context, runtime contract, PoCs, blast radius, finding confidence, exploit path, evidence gaps, and handoff notes.
   • Emit artifacts/security-review.dev.md, artifacts/security-review.appsec.md, or artifacts/security-review.exec.md only when leadership, client, or AppSec reporting is in scope.
   • Score from 100: Critical -25 (cap 25), High -15 (cap 40), Medium -8, Low -3.
   • Reuse the existing security-review.md record when pentest follows earlier design or PR security review work.
   • Deliver findings in the standard template below.

Output Template

Executive Summary

• Hacker Score: X/100 ([Critical/Vulnerable/Moderate/Hardened])
• Target Scope: [repos, URLs, mobile apps]
• Findings: [Critical/High/Medium/Low counts]

Findings Table

| ID | Title | Platform | Severity | CVSS | CWE | PoC | | --- | --- | --- | --- | --- | --- | --- | | SEC-01 | [title] | [backend|frontend|mobile] | [Critical|High|Medium|Low] | [score] | [CWE-id] | [Yes|No] |