The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): The skill presents a high-risk attack surface for indirect prompt injection. The agent is instructed to ingest untrusted data (images, metadata, tags, and annotations) from an OMERO server (references/data_access.md). There are no boundary markers or instructions to ignore embedded commands. Crucially, the skill provides powerful 'write' and 'execute' capabilities, including server-side script execution (references/scripts.md) and object deletion (references/advanced.md), without providing any sanitization or validation logic for the processed content.\n- [REMOTE_CODE_EXECUTION] (MEDIUM): The skill facilitates the creation and execution of OMERO.scripts, which run Python code server-side. While this is a standard feature of the OMERO platform, it allows the agent to trigger remote execution based on its logic, which represents a significant privilege that must be carefully controlled.\n- [EXTERNAL_DOWNLOADS] (INFO): The skill relies on omero-py and PyYAML. These are standard, well-known libraries for interacting with OMERO servers and parsing configuration files.\n- [CREDENTIALS_UNSAFE] (INFO): Code examples include placeholders for authentication credentials (username, password). The documentation correctly identifies the security risk and recommends using environment variables or external configuration files rather than hardcoding.

omero-integration