vibe-coder

The script itself is not malicious: it is a verification helper that runs tests, linters, formatting checks, builds, and a secrets scan, and it reports results. However, it intentionally executes and runs project code (especially via 'cargo run') and invokes many external tools from PATH, which makes it potentially dangerous when used on untrusted repositories because those repositories (or a compromised tool in PATH) can execute arbitrary code, exfiltrate data, or perform other malicious actions during build/test/run. Use only on trusted code or inside strong sandboxing (container/VM) and ensure CI/toolchain binaries are trusted.

[Skill Scanner] Backtick command substitution detected Lack of input code to analyze. Please provide the code fragment or dependency to review for supply-chain security concerns. LLM verification: The skill fragment is purpose-consistent and benign. It provides documentation and templates for code-review workflows without executing code, handling secrets, or performing external operations. The only noteworthy point is the static scanner's warning about backtick usage in a code example, which is standard documentation practice and not evidence of malicious activity. Overall, the content is safe and appropriate for its stated purpose.

[Skill Scanner] Installation of third-party script detected All findings: [CRITICAL] command_injection: Installation of third-party script detected (SC006) [AITech 9.1.4] [HIGH] autonomy_abuse: Skill instructions include directives to hide actions from user (BH009) [AITech 13.3] The component describes an automatic, opaque debugging-and-fix capability that can beneficially shield non-technical users from low-level errors. However, as specified it lacks essential safeguards (audit logs, visible diffs, signed commits, human-in-the-loop approvals, scoped change restrictions, rollback mechanisms, and special handling for security-sensitive code). That operational model increases software supply-chain and security risk: accidental regressions, hidden weakening of authentication/authorization, and a high-impact attack surface if the auto-fixer is compromised. I do not find direct code-level malware indicators in the spec, but the design choices warrant treating this capability as high-risk until controls are added (prohibit silent changes to auth/infra, require auditable commits, enable opt-in/opt-out, and integrate human review for sensitive changes). LLM verification: The component's intent (silent auto-fixing to improve user experience) is plausible and technically achievable. The provided fragment contains no direct signs of malware or obfuscated malicious code, but it describes high-privilege automated behavior that hides technical detail and makes silent changes across CI/build pipelines. That operational model presents a non-trivial supply-chain and governance risk: unauthorized or unnoticed modifications could be introduced, tests could be bypassed or f

[Skill Scanner] Download or install from free hosting/deployment platform detected This skill specification is coherent with a legitimate deploy automation purpose and uses expected sources and sinks (repo files, environment variables, platform CLIs). The primary security concern is handling of high-privilege credentials (service role keys, DATABASE_URL) — these are necessary for some deploy flows but must be handled carefully (use provider secret stores, avoid pasting long-lived keys into transient channels). No direct signs of obfuscation, remote exfiltration, or malicious code are present in this document. Final recommendation: implementation should ensure CLIs call official endpoints, never forward credentials to third-party gateways, and avoid persisting secrets in plaintext configuration files. Review the actual implementation before trusting it with production secrets. LLM verification: No explicit malware or obfuscated malicious code was found in the provided skill document. The primary risk is operational: it requests and would handle high-privilege credentials (service role keys, DATABASE_URL, SECRET_KEY) and executes platform CLIs to push code to production without documenting secure handling (ephemeral tokens, least privilege, authentication flow, or safe logging). If this skill were implemented as an automated agent with those permissions, it could be abused to exfiltrate