The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/test_upgrade.py uses os.system() to run shell commands constructed with f-strings (e.g., os.system(f'python "{precheck_script}" "{test_dir}"')). This pattern is vulnerable to command injection if paths are manipulated.
[DATA_EXFILTRATION]: The skill includes code in scripts/auto_fix_library.py and scripts/fix_rpc_service.py that generates JavaScript fetch() calls. While intended for Odoo RPC, this capability can be leveraged for data exfiltration if an attacker controls the endpoint.
[PROMPT_INJECTION]: The skill processes external Odoo modules (Python, XML, JS) through scripts like scripts/syntax_validator.py and scripts/upgrade_to_odoo19.py without sanitization or boundary markers.
Ingestion points: File read operations across the project directory in all major scripts.
Boundary markers: No markers are present to isolate processed data from agent instructions.
Capability inventory: Allowed access to the Bash tool and file system modification.
Sanitization: Content is processed using regex without escaping or validation of untrusted inputs.

odoo-upgrade