The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill requires the use of sudo to run DTrace, granting root-level permissions to the instrumentation process.
[COMMAND_EXECUTION]: The guide includes instructions for running various shell commands to set up the environment and execute the profiler.
[EXTERNAL_DOWNLOADS]: The setup instructions include pyenv install and pip install, which download Python interpreters and packages from remote registries. These are well-known services and do not escalate the verdict on their own.
[INDIRECT_PROMPT_INJECTION]: The skill traces programs that fetch external web content via requests and beautifulsoup4, creating a risk that malicious data captured in the trace logs could influence the agent during analysis.
Ingestion points: DTrace output logs and .trace files generated during execution.
Boundary markers: No delimiters or isolation instructions are provided for separating trace data from agent instructions.
Capability inventory: Shell access via sudo and file redirection for log creation.
Sanitization: No sanitization of the trace output is described before it is processed by the agent.

python-dtrace-profiling