python-cpu-profiling
Fail
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: HIGHPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill provides instructions for executing the 'py-spy' profiler using 'sudo', which grants the agent administrative privileges.
- File: py-spy.md
- Evidence: sudo "$(command -v py-spy)" record -o profile.svg -- python program.py
- [PRIVILEGE_ESCALATION]: The skill includes commands to tune the system for benchmarking purposes, which typically involves modifying system-level configurations and kernel parameters.
- File: pyperformance.md
- Evidence: python -m pyperf system tune
- [COMMAND_EXECUTION]: The skill relies on executing multiple shell commands to set up environments, install dependencies, and run various profiling tools.
- File: SKILL.md
- Evidence: Multiple code blocks containing 'pyenv', 'python -m venv', 'pip install', and profiler invocations like 'pyinstrument program.py'.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a target program ('program.py') that fetches and processes data from an external website (Wikipedia), which serves as an ingestion point for untrusted data.
- Ingestion points: 'program.py' (described in SKILL.md) fetches content from 'https://en.wikipedia.org/wiki/Intuitive_Machines_Nova-C'.
- Boundary markers: None identified in the skill's instructions or descriptions.
- Capability inventory: Includes shell command execution, system configuration modification (tuning), and local file creation/writing.
- Sanitization: No sanitization or validation of the external content is described.
Recommendations
- AI detected serious security threats
Audit Metadata