skills-web-dev

BENIGN in purpose alignment but HIGH RISK as an AI-agent security skill. It is coherent with its stated audit function and uses official audit tooling, but it materially equips the agent to perform security scanning and inspect sensitive configuration, so overall risk is elevated despite no evidence of malware or credential exfiltration.

No supply-chain malware is evident because this is a set of debugging/reverse-engineering instructions rather than code from a package manager dependency. The content is dual-use: it explicitly describes bypassing anti-debugging (ptrace PT_DENY_ATTACH) and bypassing iOS overlay guards by directly writing to in-process memory, plus it shows powerful in-process debugger execution (EvaluateExpression) and runtime instrumentation (DTrace objc_msgSend printing). Treat the fragment as high capability tooling; risk is misuse for unauthorized analysis rather than embedded malware in a dependency.

No evidence of overt malware or supply-chain sabotage is present in this fragment: behavior is limited to RBAC evaluation, DB queries, cache management, audit logging, and HTTP authorization gating. The dominant security concern is authorization correctness: hasPermissionOrHigher() appears to misuse async hasPermission() without await, which can cause Promise truthiness and lead to authorization bypass. Additionally, cache TTL and invalidation reliance can create temporary stale permission outcomes, increasing the chance of incorrect access decisions. Recommend fixing the async logic (await all async permission checks) and verifying consistent cache integration and event-driven invalidation coverage.

Overall, this fragment grants a highly privileged automation runtime: network fetch access to multiple domains, broad shell/Python/PowerShell execution, read/write/delete and unzip capabilities on local files, and git add/commit operations. It also strongly indicates extracting and printing text from local PDFs/EPUBs and emitting transformed stdin data to stdout, creating a clear confidentiality risk if logs/outputs are exported. While explicit malware behavior (credential theft/C2/reverse shell) is not demonstrated in this snippet alone, the capability set is consistent with potential data harvesting and exfiltration in surrounding orchestration; it should be treated as high risk and reviewed together with the consuming code paths and output/network forwarding logic.

No clear malicious payload/backdoor behavior is present in this snippet. However, it contains a significant client-side security weakness: it renders untrusted server-provided error/validation messages into SweetAlert2 using the html option without escaping or sanitization, creating a realistic DOM XSS risk. Because the module also uses a Bearer token stored in localStorage and automatically attaches it to requests, any successful XSS in the context of the application could enable token theft and account compromise. Additionally, authenticated request routing depends on caller-controlled baseUrl/endpoint and should be tightly controlled elsewhere.

No direct indicators of malware (e.g., credential theft, exfiltration, persistence) are evident in the provided fragment. However, the script performs elevated remote code execution by downloading and executing Chocolatey’s installer at runtime (iex) and then installs additional system utilities from external package feeds without pinning or integrity verification in the script. In a high-trust or tightly controlled environment, this is commonly acceptable; otherwise, it warrants review, pinning/version control, and stronger integrity assurance for the bootstrap and packages.