qa
Pass
Audited by Gen Agent Trust Hub on Aug 26, 2026
Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes the dynamic command execution syntax (
!command) to automatically retrieve context at load time. Specifically, it reads the active persona frompersonas/_active.mdand identifies the most recent file incontent/drafts/. This information is used to populate the agent's context for the QA evaluation. This behavior aligns with the skill's stated purpose and operates within the local project directory. - [INDIRECT_PROMPT_INJECTION]: The skill processes content from local draft files (
content/drafts/*.md). These files are read and included in the agent's context without explicit sanitization, which could allow content within a draft to influence the agent's evaluation behavior. - Ingestion points:
content/drafts/*.mdandpersonas/*/accounts.md(loaded via dynamic shell command). - Boundary markers: Minimal (uses
---separator). - Capability inventory: The skill documentation suggests capabilities to read files and write QA reports to
content/qa/. - Sanitization: Absent; the content of files is catenated directly into the context.
Audit Metadata