The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/agent.py defines a function run_pecmd that uses subprocess.run to call an external forensic binary, PECmd.exe. The command is constructed as a list, which mitigates shell injection risks, but the skill relies on the presence of this external tool in the environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes Windows Prefetch files, which are untrusted forensic artifacts. Maliciously crafted executable names or internal metadata within these files could potentially influence an agent's behavior if it interprets the extracted strings as instructions during the summarization or analysis phase.
Ingestion points: scripts/agent.py reads filenames using glob.glob and binary content using struct.unpack; SKILL.md processes files via shell commands like ls and grep.
Boundary markers: None. There are no delimiters or instructions to ignore embedded content when parsing forensic strings.
Capability inventory: The skill uses subprocess.run in scripts/agent.py and standard shell commands (mount, cp, sha256sum) in the SKILL.md workflows.
Sanitization: The script performs basic UTF-16 decoding (decode('utf-16-le', errors='replace')) but lacks specific sanitization to prevent the interpretation of extracted strings as prompt instructions.
[EXTERNAL_DOWNLOADS]: The SKILL.md file specifies the installation of the prefetch package via pip and references the official repository for Eric Zimmerman's forensic tools (https://ericzimmerman.github.io/) for downloading PECmd. These are standard, well-known resources in the digital forensics community.

analyzing-prefetch-files-for-execution-history