The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from upgrade logs provided by the user.
Ingestion points: The analyze_upgrade_log tool in SKILL.md accepts log_text directly from user-provided logs.
Boundary markers: There are no explicit delimiters or instructions to ignore commands within the log input.
Capability inventory: The skill has read access to source code via scan_addons_source and data-writing capabilities through a gated write workflow.
Sanitization: The instructions do not define sanitization or filtering for the ingested log data.

odoo-migration-copilot