The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/crash_trace.py script executes the rr debugger tool using the subprocess module. The implementation demonstrates high security awareness, utilizing multiple hardening layers to mitigate risks associated with debugging potentially untrusted binaries: Disables GDB init files (-nx) and all auto-load mechanisms (scripts, Python) via command-line flags; Sets a neutral environment by stripping most environment variables and using a temporary home directory; Executes in a neutral, empty working directory to prevent side-loading.
[DYNAMIC_EXECUTION]: The test file scripts/test_crash_trace.py uses importlib.util to load the crash_trace.py module for unit testing. This is a standard Python idiom for testing local modules.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from an external tool (rr replay). While it lacks explicit boundary markers for the captured output, it is intended for debugging scenarios where the user interacts with known codebases. Ingestion point: scripts/crash_trace.py reads debugger output. Boundary markers: Absent. Capability inventory: Shell command execution via subprocess for the debugger. Sanitization: Basic UTF-8 decoding with replacement for invalid characters.

rr-debugger