headless-ghidra-discovery
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from target analysis artifacts to generate metadata updates.
- Ingestion points: The skill reads various YAML files from the
artifacts/directory, includingpipeline-state.yaml,runtime/hotpaths/call-chain.yaml, andthird-party/identified.yaml, which contain results from previous analysis phases. - Boundary markers: No specific boundary markers or instructions to ignore embedded commands are present in the inputs, though the content is processed as structured YAML.
- Capability inventory: The skill utilizes the
ghidra-agent-clito modify the Ghidra project state, including renaming functions and applying signatures. - Sanitization: The workflow incorporates mandatory validation (
metadata validate,hotpath validate) and verification (verify-renames,verify-signatures) commands that check metadata integrity before and after application, mitigating the risk of processing malicious inputs. - [COMMAND_EXECUTION]: The skill invokes the
ghidra-agent-clitool to perform workspace management, metadata enrichment, and Ghidra project mutations. These commands are vendor-provided resources essential to the skill's intended functionality within the reverse engineering pipeline.
Audit Metadata