code-refactoring-tech-debt
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection through the direct interpolation of user-supplied content.
- Ingestion points: The skill ingests untrusted data via the
$ARGUMENTSvariable located in theSKILL.mdfile under the Requirements section. - Boundary markers: The variable is placed directly into the instruction flow without the use of boundary markers, delimiters (such as XML tags or triple quotes), or specific instructions for the agent to treat this data as distinct from the skill's core logic.
- Capability inventory: While the skill does not explicitly request high-privilege tools in its frontmatter, it is designed for codebase analysis, which typically grants the agent access to the project's file system and terminal tools within its operating environment.
- Sanitization: There is no evidence of filtering, escaping, or validation performed on the
$ARGUMENTSinput to prevent malicious instructions from influencing the agent's behavior.
Audit Metadata