The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the use of sudo for administrative tasks, including installing packages through system managers (apt, dnf, pacman), installing software compiled from source code (sudo make install), and modifying kernel parameters to enable performance monitoring features (sudo sysctl kernel.perf_event_paranoid=1).
[EXTERNAL_DOWNLOADS]: The skill fetches source code from a remote GitHub repository (https://github.com/rr-debugger/rr) and instructs the user to compile and install it. It also relies on external package repositories for tool installation.
[COMMAND_EXECUTION]: The skill relies extensively on shell commands for environment setup, tool installation, and the dynamic execution of target binaries under analysis.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection as it is designed to process untrusted external data.
Ingestion points: Untrusted data enters the agent's context through crash_input files, which are processed as standard input or command-line arguments by the debugging tools.
Boundary markers: There are no explicit boundary markers or warnings to prevent the agent from interpreting instructions potentially embedded within the crash samples.
Capability inventory: The skill utilizes powerful tools including gdb, rr, afl-tmin, and the execution of the target binary (./target), all of which operate on the untrusted input.
Sanitization: No sanitization or validation of the input data is described before it is passed to the execution environment.

re-crash-triage