offensive-windows-mitigations
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for auditing external infrastructure, binaries, and source code, which serves as an ingestion surface for untrusted data. \n
- Ingestion points: Operational workflow and discovery checklist in
SKILL.mdmention identifying entrypoints, trust boundaries, and performing source/binary/bundle inspection. \n - Boundary markers: Absent; there are no instructions to use delimiters or specific ignore-tags when processing audited content. \n
- Capability inventory: The methodology involves static taint analysis, binary disassembly, decompilation, and shell command execution. \n
- Sanitization: No sanitization or escaping of external content is specified before the agent processes or interprets the data. \n- [COMMAND_EXECUTION]: The skill instructs the agent to run local commands as part of the verification process. \n
- Evidence: The remediation section of the checklist in
SKILL.mdexplicitly directs the user/agent to executenpm testand utilize toolchain build and test suites to verify that flaws are resolved.
Audit Metadata