mizchi-blog-style
Pass
Audited by Gen Agent Trust Hub on Oct 12, 2026
Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill references absolute local file paths that reveal the system username 'mz'.
- Evidence:
/Users/mz/.claude/skills/mizchi-blog-style/SKILL.mdand~/ghq/github.com/mizchi/zenn/CLAUDE.mdinSKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided draft articles, creating a potential surface for indirect prompt injection when these drafts are passed to a subagent for review.
- Ingestion points: The subagent dispatch template takes a file path
<記事ファイルのパス>as input inSKILL.md. - Boundary markers: The prompt template lacks explicit delimiters or instructions to ignore embedded commands within the draft article.
- Capability inventory: The subagent is instructed to
Readspecific files based on the prompt instructions. - Sanitization: There is no evidence of sanitization or filtering applied to the draft content before it is analyzed by the subagent.
Audit Metadata