The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from binary files and uses it to perform actions like renaming functions and adding documentation, creating an attack surface for indirect prompt injection.
Ingestion points: The skill reads procedure_pseudo_code, procedure_assembly, procedure_callers, procedure_callees, and arbitrary strings referenced by the binary (SKILL.md).
Boundary markers: There are no delimiters or instructions to ignore embedded commands within the binary data being analyzed, which could allow malicious strings in a binary to influence the agent's behavior.
Capability inventory: The skill utilizes rename_function and add_note tools to update the environment based on potentially attacker-controlled input.
Sanitization: No sanitization or validation is applied to the data extracted from the binary before it is processed or used in tool calls.

decompile