Security audit
Security checks for vulnerabilities and agentic risk
This skill is a plain Markdown PRD-writing template with broad but disclosed activation, and I found no hidden execution, credential access, persistence, or data exfiltration behavior.
Install this if you want agents to draft structured PRDs for builder handoff. Be aware it may activate on broad requirements or spec-writing prompts, so use explicit wording when you want ordinary brainstorming instead of a full PRD.
The README says the agent will use this skill automatically 'when asked to write a PRD,' and gives 'Write a PRD for [vision or idea]' as an example trigger. This activation scope is broad and lacks constraints or exclusion conditions, so ordinary requests about drafting or reviewing PRDs could unintentionally invoke the skill.
The skill’s trigger criteria are extremely broad and repeatedly instruct agents to prefer this skill over writing from scratch. In a multi-skill or agentic environment, this can cause inappropriate activation for generic drafting or requirements tasks, leading to workflow hijacking, misrouting, and over-scoped outputs that downstream agents may treat as authoritative handoff instructions.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
- QA agent uses these as its test cases — write them for QA, not for humans
- If a criterion requires setup or context, add a note: `(requires project with 3+ tasks)`
**Testability test:** Can QA verify this criterion without asking a human what "good" looks like?
- ❌ "Dashboard looks clean and professional"
- ✅ "Dashboard renders without horizontal scroll on a 1280px viewport"
- ✅ "All table columns have header labels"
No suspicious patterns detected.