empirical-prompt-tuning
Pass
Audited by Gen Agent Trust Hub on Apr 27, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and execute external prompts via subagents, creating an indirect prompt injection surface.
- Ingestion points: The skill reads a 'Target prompt' and scenario descriptions from the environment or user input, as specified in the 'Workflow' and 'Subagent invocation contract' sections of
SKILL.mdandSKILL-ja.md. - Boundary markers: The instructions use Markdown headers (e.g.,
## Target prompt,## Scenario) to delimit external content, which provides some structural separation but does not prevent intentional injection. - Capability inventory: The skill uses the
Tasktool to dispatch autonomous subagents and involves file-system operations to apply prompt 'diffs' (Workflow Step 5). - Sanitization: There is no evidence of automated sanitization, filtering, or validation of the input prompt content before it is passed to the subagent.
Audit Metadata