The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external code from sources/ and projects/ directories. This untrusted content is ingested into the agent context without explicit boundary markers or delimiters. Combined with the skill's capabilities to perform file writes and execute Docker commands, this creates an attack surface for indirect prompt injection.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute high-privilege shell commands, including docker, docker-compose, psql, and curl. While these are necessary for the skill's DevOps purpose, they represent a significant capability surface.
[CREDENTIALS_UNSAFE]: The skill manages sensitive information such as database credentials, Odoo master passwords, and DOCKERHUB_TOKEN via environment variables and .env files. While the skill encourages security hardening (e.g., list_db = False), the exposure of these credentials to the agent context is a noted risk factor.

odoo-docker