The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection because it is designed to ingest and analyze untrusted external data.
Ingestion points: The skill reads binary files via open_binary, crash dumps via inspect_recorded_crash, and runtime process memory/traces via observe_native_calls and capture_process_scenario.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions that might be embedded in the binary data or process memory.
Capability inventory: The skill has access to powerful debugging and instrumentation tools including Frida (instrument_with_frida), GDB (gdb_console), and Rizin (rizin_debug_command), which allow for deep process interaction.
Sanitization: There are no explicit sanitization or filtering steps described for the content extracted from memory or files before it is processed by the agent.
[COMMAND_EXECUTION]: The skill provides the agent with capabilities to execute commands within debugging environments (GDB, Rizin) and instrument live processes with Frida. These are powerful capabilities that allow low-level system interaction as part of its primary forensic purpose.

rea-performance-crash-triage