universal-reverse-engineer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt explicitly instructs extracting and logging plaintext strings (including hooking decryption functions) and to list "hardcoded creds" / IOCs in reports, which requires the agent to output any secrets found verbatim.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The content is a dual‑use reverse‑engineering toolkit that explicitly documents and automates techniques for credential extraction (crypto/SharedPreferences/Keychain hooks), SSL‑pinning and root/debugger bypasses, repackaging/injection (Frida‑gadget/smali patches), exposing remote access (frida‑server listening, gadget injection), system‑level compromises (mounting /system, disabling SIP/AMFI, iptables redirects), and anti‑analysis evasion—patterns which enable backdoors, remote code execution, and data exfiltration if abused.

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

• Third-party content exposure detected (high risk: 0.78). The skill’s runtime workflow ingests outsider-authored free text when it decodes user-supplied APK/ZIP binaries (e.g., scripts/triage.sh → unzip -l/apktool d and scripts/strings-extract.py → reads ZIP contents and extracts strings from embedded classes.dex/AndroidManifest.xml), and those embedded text artifacts are authored by the APK’s publisher (an outsider) and become LLM-readable via extracted strings/decompiled output.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs operators to fetch and install a remote native payload (frida-gadget) via "wget https://github.com/frida/frida/releases/download/16.x.x/frida-gadget-16.x.x-android-arm64.so.xz" and then load it into APKs (which executes remote code at runtime), so this external URL is a runtime dependency that results in execution of downloaded code.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill instructs the agent to perform state-changing, privileged actions (patch binaries, rewrite /proc/self/status, run frida-server, use LD_PRELOAD/ScyllaHide, etc.) that bypass security mechanisms and modify system files/behavior, which can compromise the host.