Security audit
Security checks for vulnerabilities and agentic risk
This meal-planning skill stores household food preferences locally and appears consistent with that purpose, with no evidence of hidden execution or data sharing.
Install only if you are comfortable keeping meal-planning memory in local plaintext markdown files. Consider using nicknames or coarse age groups and storing only dietary or allergy details needed for planning; delete ~/meal-planner/ to remove the stored data.
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
integration: pending | done | declined
## Household
<!-- Who you're cooking for -->
### Members
| Name | Age Group | Notes |
Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.
<!-- vegetarian, pescatarian, keto, etc. — whole household or per person -->
### Universal Dislikes
<!-- Foods NO ONE in household wants -->
## Preferences
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.
## Common Traps
- Planning without checking inventory → duplicate purchases, waste
- Overambitious meal plans → exhaustion, ordering takeout
- Ignoring prep time → not just cook time, total time matters
- Same proteins all week → meal fatigue, nutrition gaps
The template explicitly directs persistent storage of household names, age groups, and sensitive dietary and allergy information, including a child and a life-threatening allergy, without any privacy guidance, minimization, consent, retention, or access-control considerations. In a meal-planning context this data is operationally useful, but it is still sensitive household health/profile data that could be exposed through local compromise, sync, backups, or unintended sharing.
The activation condition triggers whenever ~/meal-planner/ does not exist, but it does not narrowly constrain when the skill should engage in conversation. That can cause the agent to initiate setup outside a clearly user-requested meal-planning context, increasing the chance of unwanted persistence, over-collection of personal preferences, or intrusive behavior.
The prompts ask whether the assistant should help 'whenever you're thinking about food' and whether it should proactively remind the user, which expands the skill beyond a narrow task scope. In practice, this could let the skill opportunistically engage during ordinary conversation and encourage persistent monitoring-like behavior that the user may not expect.
No suspicious patterns detected.