prompt-eval
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves ingesting and acting upon untrusted external prompts, system instructions, and agent configurations, which creates a vulnerability surface for indirect injection attacks.\n
- Ingestion points: In
SKILL.md(Sections 1 and 3), the agent is instructed to read 'System Prompt', 'Skill/rule files', 'tool descriptions', 'request templates', and 'running configurations' for evaluation.\n - Boundary markers: The skill does not provide specific instructions to use delimiters or 'ignore embedded instructions' markers when processing the data from the prompts being audited.\n
- Capability inventory: The skill possesses the capability to update and modify files, as specified in
SKILL.md(Section 4: '提出最小修正'), increasing the potential impact if a processed prompt contains malicious modification commands.\n - Sanitization: There are no instructions for sanitizing or validating the ingested content before it is processed or reflected back into the workspace.
Audit Metadata