The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted binary samples using tools like strings, readelf, and objdump. The output of these tools is then analyzed by the agent.
Ingestion points: Binary files (sample.bin) and their metadata.
Boundary markers: Absent. There are no instructions for the agent to treat the output of analysis tools (like extracted strings) as data rather than instructions.
Capability inventory: The skill utilizes shell commands (file, strings, objdump, etc.) and internal Python execution.
Sanitization: None. The skill presents raw output from binary analysis tools directly to the agent's context.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute various system utilities (file, sha256sum, strings, objdump, readelf, xxd) to analyze local files.
[DYNAMIC_EXECUTION]: The skill includes a pattern for generating and executing a Python script at runtime via a shell heredoc to calculate file entropy. Although the provided script is static and performs benign mathematical operations, it represents a dynamic code execution vector.

re-triage