aws-penetration-testing

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes commands and examples that embed API keys, temporary credentials, and secrets verbatim (e.g., CLI args, export AWS_ACCESS_KEY_ID/SECRET, enumerate-iam --access-key/--secret-key), which requires the agent to handle and output secret values directly, creating an exfiltration risk.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These URLs are high-risk because they include direct access to AWS metadata endpoints (used for SSRF/credential theft), public S3 bucket endpoints and bucket-indexing services (which can host arbitrary/malicious binaries), proxying of metadata, and several GitHub repos and tools that—while legitimate for pentesting—can distribute executables or be abused to deploy malware.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This content explicitly instructs how to steal credentials and secrets (SSRF to IMDS, container/env reads, Secrets Manager, STS/STS-derived tokens), exfiltrate data (S3 sync, EBS/NTDS extraction), escalate privileges and create persistent admin access (create-access-key, attach policies, CreateLoginProfile, PassRole/RunInstances), install backdoors and remote code execution (update Lambda code, build/push backdoored containers, Lambda/API Gateway backdoors), and evade detection (disable CloudTrail, user-agent obfuscation), i.e. it provides step‑by‑step malicious techniques for unauthorized access, persistence, and cover‑up.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). The skill's core workflow (SKILL.md) explicitly instructs fetching and inspecting untrusted third‑party content — e.g., git clone of public GitHub repos, using aws s3 sync / public bucket URLs (including https://buckets.grayhatwarfare.com), and aws lambda get-function plus wget to download Lambda code — and those retrieved files/outputs are read and used to drive escalation and follow-up actions, so untrusted content can materially influence agent behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill instructs runtime installation and execution of external tooling (e.g., "git clone https://github.com/RhinoSecurityLabs/pacu"), which fetches and executes remote code and is listed as an essential dependency, so this URL represents a high-risk runtime external dependency.