hardware-security

Pass

Audited by Gen Agent Trust Hub on Oct 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute powerful hardware-interfacing CLI tools, including openocd for JTAG debugging, flashrom for SPI flash manipulation, and sigrok-cli for bus analysis. These tools possess the capability to read from and write to device memory and firmware.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external, potentially untrusted data sources such as firmware dumps and logic analyzer captures.
  • Ingestion points: Processes files using binwalk (SKILL.md), sigrok-cli (SKILL.md), and flashrom (SKILL.md).
  • Boundary markers: The instructions lack explicit boundary markers or warnings to the agent to ignore instructions that might be embedded within analyzed firmware or data streams.
  • Capability inventory: The skill is granted Bash(*), Read, Write, and Edit tool permissions, which could be abused if an attacker hides instructions in a firmware image.
  • Sanitization: No sanitization or validation of the content of firmware or bus captures is performed before processing.
  • [METADATA_POISONING]: There is a discrepancy in author information between the SKILL.md frontmatter, which lists babysitter-sdk, and the metadata.json file, which lists majiayu000. This suggests the skill may have been aggregated or forked without updating metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 10, 2026, 06:32 PM