blender-stop-condition-decider

Pass

Audited by Gen Agent Trust Hub on Oct 24, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources such as blender-recipe-decomposer and acceptance-criteria-builder (specifically the 'Single recipe scope' and 'Measurable acceptance rows') without implementing safety boundaries or sanitization steps. This creates a potential for malicious instructions embedded in that data to influence the agent's behavior during the handoff process.
  • Ingestion points: Defined in SKILL.md under the 'Input schema' section, which consumes scope and acceptance criteria from external skills.
  • Boundary markers: Absent. There are no instructions to wrap input data in delimiters or warnings to disregard embedded commands.
  • Capability inventory: The skill has the capability to determine and trigger the 'handoff' to subsequent skills in the catalog, allowing it to control the workflow sequence.
  • Sanitization: Absent. The operating procedure does not include steps for validating or escaping the content of the ingested data before it is formatted into the stop condition output.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 24, 2026, 07:02 PM