The Agent Skills Directory

[METADATA_POISONING]: The skill metadata in SKILL.md attributes the author as 'mahipal', which differs from the LICENSE file and internal context identifying 'mukul975' as the primary author. Additionally, the license contains a future-dated copyright (2026).
[EXTERNAL_DOWNLOADS]: The skill documentation recommends installing the drozer package via pip. Drozer is a reputable and widely used security assessment framework for the Android platform.
[COMMAND_EXECUTION]: The scripts/agent.py script utilizes the subprocess.run library to interact with host-level binaries such as adb and drozer to perform its security analysis functions.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses externally supplied manifest files and package names without adequate sanitization, potentially leading to command injection if these inputs are malicious.
Ingestion points: Package names provided to scripts/agent.py and AndroidManifest.xml files processed by scripts/process.py.
Boundary markers: No delimiters or safety instructions are used to separate untrusted inputs from the execution logic.
Capability inventory: The skill has the ability to execute shell-level commands through subprocess.run calls to system utilities and security frameworks.
Sanitization: There is no evidence of input validation or escaping for parameters passed to the Drozer and ADB command strings.

testing-android-intents-for-vulnerabilities