skills/trothbyte/low-level-skills-trothbyte/reverse-engineering-ghidra-agent-automation/Gen Agent Trust Hub
reverse-engineering-ghidra-agent-automation
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted binary data (e.g., target programs, ROMs), creating a surface for indirect prompt injection if the agent interprets embedded binary content as instructions.
- Ingestion points: Untrusted binary data enters the agent context via
pyghidra.open_program(path)inexamples/good/analysis_loop.py,examples/good/rebase_and_patch.py, andexamples/bad/confident_verdict.py. - Boundary markers: The skill explicitly instructs the agent to use status markers (
KNOWN,INFERRED,VERIFIED,UNVERIFIED) to label and gate its findings, which helps differentiate between confirmed facts and potentially untrusted data observations. - Capability inventory: The skill utilizes tools to perform file system writes (
json.dumpfor triage reports) and memory modifications (Memory.setBytefor patching) based on analysis results. - Sanitization: The instructions mandate logic-level verification such as read-back checks for memory patches and multi-point falsification checks (hashes, headers) for program identity to ensure the agent is not misled by malicious inputs.
Audit Metadata