The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/agent.py script executes local AFL++ binaries including afl-cc, afl-cmin, afl-tmin, and afl-whatsup via subprocess.run(). These commands are used to facilitate the fuzzing workflow and are invoked with controlled argument lists.
[EXTERNAL_DOWNLOADS]: The references/api-reference.md documentation provides installation instructions that reference the official AFL++ GitHub repository and standard system package managers. These references target well-known and trusted sources for security auditing software.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data produced by the fuzzing process, such as campaign metrics and crash file metadata, which could theoretically be manipulated by a malicious target program.
Ingestion points: The script parses the fuzzer_stats file and enumerates filenames in the crashes directory within scripts/agent.py.
Boundary markers: None; the extracted metadata is printed directly in the audit report.
Capability inventory: The skill includes capabilities for file system access and local command execution via sub-processes.
Sanitization: There is no specific sanitization of the fuzzer-generated strings before they are displayed, although the script primarily extracts structured numerical or technical data.

performing-fuzzing-with-aflplusplus