The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including memory dumps, disk images, and suspicious files, which represents an attack surface for indirect prompt injection.
Ingestion points: Files processed by file, exiftool, strings, binwalk, and other forensic utilities (documented across all files).
Boundary markers: None identified; the agent is instructed to process the raw output of forensic tools directly.
Capability inventory: The skill has access to Bash, Read, Write, and Grep tools, allowing it to execute system commands and modify the file system.
Sanitization: No explicit sanitization or filtering of tool output is performed before processing by the LLM.
[DYNAMIC_EXECUTION]: The skill includes Python code snippets intended for execution by the agent to perform specific forensic tasks like LSB extraction.
Evidence: reference/steganography.md contains Python scripts using the PIL library for image processing and standard file operations to detect appended data.
Capability Tier: The scripts are simple templates for known forensic tasks and do not incorporate untrusted external input into their logic.

forensics