golang-benchmark
Warn
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: MEDIUMPRIVILEGE_ESCALATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PRIVILEGE_ESCALATION]: The skill provides instructions for system-level hardware tuning using
sudoto modify settings such as the CPU scaling governor, Turbo Boost, and SMT (Hyper-Threading) on dedicated CI runners. While documented for optimization, these are administrative operations with elevated privileges.\n- [COMMAND_EXECUTION]: The skill utilizes numerous shell commands to run benchmarks, generate profiles, and analyze execution traces via the Go toolchain and third-party utilities.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (benchmark outputs, pprof profiles, and execution traces), which creates a potential surface for indirect prompt injection if the inputs originate from untrusted environments.\n - Ingestion points: Reads benchmark results (
bench.txt), heap/CPU profiles (cpu.prof,mem.prof), and execution traces (trace.out).\n - Boundary markers: None identified during the parsing of tool output.\n
- Capability inventory: Shell command execution for benchmarking and profiling utilities.\n
- Sanitization: No documented sanitization for ingested performance data.\n- [EXTERNAL_DOWNLOADS]: Recommends installing performance tools from well-known sources, including official Go sub-repositories and established community projects on GitHub.
Audit Metadata