The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill demonstrates the use of Frida's CModule API to compile C source code strings directly into the memory of a target process at runtime for performance-critical instrumentation.
Evidence: references/hooks-catalog.md shows new CModule() usage for high-performance instruction tracing and function patching.
[COMMAND_EXECUTION]: The skill utilizes shell commands and the adb utility to set up the instrumentation environment and interact with processes.
Evidence: SKILL.md includes commands for listing processes (frida-ps), spawning binaries (frida -f), and pushing the instrumentation server to Android devices via adb push.
[DATA_EXFILTRATION]: Instructions and scripts are provided to monitor and capture sensitive data such as passwords, cryptographic keys, and network payloads from external processes.
Evidence: references/hooks-catalog.md contains a hook for the read function that uses a regular expression to search for sensitive strings like 'password', 'secret', 'key', or 'token' in data buffers.
[PRIVILEGE_ESCALATION]: The skill includes commands to modify file permissions on a device to allow for the execution of the instrumentation server.
Evidence: SKILL.md provides an adb shell chmod 755 command for the frida-server binary.
[EXTERNAL_DOWNLOADS]: The skill references external binaries hosted on GitHub for the instrumentation toolkit's server-side component.
Evidence: SKILL.md and references/android-ios.md point to the official GitHub releases page for frida-server downloads.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external processes, which could potentially contain malicious instructions targeting the agent.
Ingestion points: Data is read from process memory and system calls using Interceptor.attach and Memory.readByteArray in SKILL.md and references/hooks-catalog.md.
Boundary markers: Absent.
Capability inventory: Process injection, memory modification, and shell command execution are detailed across all scripts in the skill.
Sanitization: Absent. Content from hooked processes is logged directly to the console without filtering or escaping.

frida