The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access behaviors were detected. The skill maintains a clear focus on its intended reverse-engineering purpose using legitimate vendor tools.
[COMMAND_EXECUTION]: The skill utilizes the ghidra-agent-cli utility, which is a vendor-provided tool from ByteLandTechnology. Command execution is limited to workspace setup and binary metadata inspection. A safety constraint explicitly forbids the agent from creating or running new Ghidra scripts without user approval.
[PROMPT_INJECTION]: The skill defines an ingestion point for untrusted data in the form of user-provided binaries. This constitutes an indirect prompt injection surface.
Ingestion points: Identified in SKILL.md as 'User-provided binary or archive path'.
Boundary markers: Present in planning-brief.md through 'Non-Negotiable Reverse-Engineering Constraints' and explicit 'Validation expectations'.
Capability inventory: Defined in SKILL.md as a subset of ghidra-agent-cli workspace and inspection commands.
Sanitization: Not explicitly specified in the provided markdown instructions.

headless-ghidra-intake