The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze untrusted external data, including malware samples and memory dumps, which constitutes a potential surface for indirect prompt injection.
Ingestion points: The agent recursively scans files from directories such as /mnt/evidence/collected_samples/ and /mnt/quarantine/ as documented in scripts/agent.py and SKILL.md.
Boundary markers: There are no explicit instructions or markers provided to the agent to ignore potential instructions embedded within the files being scanned.
Capability inventory: The skill utilizes file system access, directory traversal, and SHA256 hashing to process these files.
Sanitization: Results in scripts/agent.py are sanitized by hex-encoding binary data or truncating strings to 64 characters with replacement for non-UTF8 characters.
[EXTERNAL_DOWNLOADS]: The skill downloads security tools and YARA rule sets from external community repositories.
Fetches the yarGen automated rule generation tool and its databases from a public GitHub repository.
Downloads community signature bases and rule sets from repositories managed by ReversingLabs and the wider security community.
[COMMAND_EXECUTION]: Instructions include the installation of system packages and the execution of external Python tools.
Uses sudo apt install for installing the YARA engine on Linux systems.
Executes yarGen.py for rule generation and database updates, which involves running a downloaded script with system-level Python.
[DYNAMIC_EXECUTION]: The skill compiles YARA rules at runtime from source strings or external files.
scripts/agent.py and the documentation in SKILL.md use yara.compile to dynamically translate YARA rule logic into the engine's internal format for scanning.

performing-threat-hunting-with-yara-rules