The Agent Skills Directory

[DYNAMIC_EXECUTION]: The generate_validation_report.py script constructs HTML reports through direct string interpolation of results stored in a JSON file. Metadata such as table names or column names, if sourced from a compromised or malicious database, could contain script tags leading to Cross-Site Scripting (XSS) when the generated HTML file is rendered in a browser.
[DYNAMIC_EXECUTION]: The configure_validation_rules.py script supports the definition of "custom" validation rules consisting of arbitrary SQL queries. These queries are stored in a configuration file for later execution, providing a mechanism for arbitrary code (SQL) execution within the database context if the configuration is modified by an untrusted party or the user inputs malicious queries.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: Data and metadata (table/column names) are ingested from databases via psql or mysql CLI tools as described in the instructions within SKILL.md.
Boundary markers: The instructions lack specific delimiters or instructions for the agent to ignore potential instructions embedded within the ingested database content.
Capability inventory: The skill possesses Bash(psql:*), Bash(mysql:*), and Write capabilities, enabling database schema modification and local file system writes.
Sanitization: There is no evidence of sanitization or escaping of ingested database metadata or content before its inclusion in reports or potential interpolation into agent prompts.
[COMMAND_EXECUTION]: The skill requires and utilizes database CLI tools with administrative-level permissions (ALTER TABLE, CREATE FUNCTION) to implement integrity constraints. While aligned with the stated primary purpose, this grants the agent extensive control over the target database's structure and configuration.

validating-database-integrity