skills/trailofbits/skills/libafl/Gen Agent Trust Hub

libafl

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the official Rust installer from sh.rustup.rs and the LLVM installation script from apt.llvm.org.
  • [REMOTE_CODE_EXECUTION]: Executes the Rust installation script by piping the download directly to the shell (curl | sh).
  • [COMMAND_EXECUTION]: Instructs the agent to perform compilation tasks using clang++, manage Rust projects with cargo, and execute fuzzer binaries.
  • [PRIVILEGE_ESCALATION]: Uses sudo to execute the LLVM installation script, which is a standard requirement for system-level package installation on Linux.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied C++ and Rust code files (harness.cc, main.cc, Cargo.toml) which are subsequently used in compilation and execution commands.
  • Ingestion points: The agent reads harness.cc, main.cc, and project configuration files.
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore embedded instructions in the ingested code.
  • Capability inventory: The skill utilizes clang++, cargo, and shell execution to build and run targets.
  • Sanitization: No sanitization or validation of the input code is performed before it is passed to the compiler or executed.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 04:52 AM
Security Audit — agent-trust-hub — libafl