hotfix
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data via the
$ARGUMENTSvariable, which handles bug descriptions, issue links, or error messages. This content is interpolated directly into a task for the@codebase-review-agent. - Ingestion points: The
$ARGUMENTSvariable inSKILL.md(Phase 1) is the entry point for external data. - Boundary markers: The user-provided content is wrapped in
<bug_description>tags. While these tags act as delimiters, they do not prevent an attacker from attempting to escape the tags or embed instructions that could override the sub-agent's primary task. - Capability inventory: The skill possesses significant capabilities, including modifying the codebase, executing
gitcommands, deleting directories (docs/hotfix-review/), and orchestrating multiple sub-agents. - Sanitization: There is no evidence of input validation, filtering, or escaping performed on the
$ARGUMENTSbefore they are processed by the agent pipeline.
Audit Metadata