T09 · Insecure Skill Coding Practices
- Location
SKILL.md:15- Finding
Arbitrary Odoo Server Override Can Expose Fallback Credentials
- Content
View full analysis
- Remediation
View remediation
Security audit
Security checks for vulnerabilities and agentic risk
This Odoo skill is coherent but needs Review because it can broadly modify or delete live business data and may reuse saved credentials with a changed server URL.
Install only if you are comfortable giving the agent Odoo account authority. Use a dedicated least-privilege API key, restrict use to trusted Odoo URLs, verify the resolved URL/database/model/record count before create/write/unlink, and avoid production mutations unless you add explicit confirmation and preview steps.
SKILL.md:15Arbitrary Odoo Server Override Can Expose Fallback Credentials
The README explicitly promotes destructive capabilities such as deleting test partners and mass-updating records, but it does not state that write/delete actions require explicit confirmation or operator review before execution. In a generic Odoo CRUD skill that can target any model and supports dynamic instance/database switching, this omission increases the chance of accidental high-impact changes to production data from a simple user prompt or context mistake.
The skill advertises generic CRUD on any Odoo model and dynamic switching between instances/databases, but does not prominently warn that these operations may affect live production business data. Because Odoo models can include accounting, CRM, inventory, HR, and metadata objects, broad write/delete capability substantially increases the risk of damaging critical records through misuse, misunderstanding, or prompt-induced errors.
The skill explicitly documents unlink operations, including direct partner deletion examples, but provides no requirement for explicit confirmation, dry-run preview, or safeguards for destructive actions. In a production ERP context, accidental or overly broad deletions can cause irreversible business data loss and operational disruption even when the underlying user is authenticated and authorized.
The README presents the skill entirely in French and explicitly says the user intents shown are 'en français', which may imply a fixed language expectation. There is no note that other languages are supported or that French is a documented, justified locale constraint.
The user example prompts are written in French throughout multiple sections, with no indication that other languages are supported or that French is optional. This can be interpreted as a language/locale preference embedded in the skill documentation without user opt-in.