apk-reverse

Pass

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Python subprocess module to execute external tools necessary for reverse engineering. This includes:
  • adb for device interaction (found in coldstart.py, install_test.py, preflight.py, run_probe.py, spawn_patch_detach.py).
  • java, javac, jar, keytool, zipalign, and apksigner for APK repacking and signing (found in repack.py, smtool.py, rasc_build.py).
  • git and cargo for building Rust-based indexers (rasc_build.py).
  • Mitigation: The skill employs a centralized validation module (scripts/device_shell.py) which enforces strict regex-based grammar for package names, component paths, and process IDs. It also uses POSIX single-quote escaping for shell commands to prevent arbitrary command injection through malicious inputs.
  • [EXTERNAL_DOWNLOADS]: Several scripts facilitate the acquisition of specialized reverse engineering tools:
  • scripts/rasc_build.py clones a Rust-based DEX indexer from a GitHub repository (MG1937/ASC).
  • scripts/ete_capture.py provides build instructions that involve cloning OpenCSD and the Linux kernel source for hardware tracing.
  • scripts/doctor.py and scripts/capabilities.py check for the presence of numerous external binaries and provide download URLs for missing components (e.g., dexsql, jadx, apktool).
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and parse untrusted data from target Android devices and applications:
  • scripts/native_crash.py parses potentially attacker-influenced logcat captures and tombstones.
  • scripts/stalker_report.py parses instruction-level traces.
  • scripts/protobuf_decode_raw.py decodes arbitrary binary/hex payloads.
  • Assessment: While these scripts handle external data, they use deterministic parsing logic (regex, struct unpacking) and do not pass this data back into the LLM context in a way that overrides system behavior. The risk is assessed as low due to the focus on structural analysis and the inclusion of data-handling safety valves.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 18, 2026, 03:42 PM