classify-parity-differences
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes JSON files from the
.replacement/differences/directory to summarize system parity differences. This creates a surface where an attacker who can influence the creation of these files (e.g., through the system being tested) could manipulate the agent's summary reports and the subsequent manifest updates. \n - Ingestion points: The
scripts/summarize-differences.pyscript reads and parses all files matching*.jsonin the user-provided directory (.replacement/differences). \n - Boundary markers: Absent. The instructions do not specify a method for the agent to verify the integrity or origin of the JSON difference files before processing them. \n
- Capability inventory: The skill calls
python3to execute the localsummarize-differences.pyand uses the output to invokeupdate-manifest.py. \n - Sanitization: The Python script performs basic JSON structure validation but does not perform semantic validation to prevent an attacker from injecting numerous records to influence the 'blocking' counts or evidence fields. \n- [COMMAND_EXECUTION]: The skill instructs the agent to execute local scripts (
summarize-differences.pyandupdate-manifest.py) to automate system migration workflows. While these are part of the intended functionality, they involve shell execution patterns that handle data derived from local files.
Audit Metadata