The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements dynamic path resolution to locate and load external Python modules and execute scripts. In reference/mysql.md, it uses a shell find command to locate the mysql-intelligence directory and suggests adding it to the Python search path using sys.path.insert(0, ...). This allows for the execution of code from directories discovered at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and sampling data from external databases (PostgreSQL, MySQL, SQLite, MongoDB), creating a significant attack surface for indirect prompt injection.
Ingestion points: The skill uses schema_extractor.py and document sampling techniques (e.g., MongoDB aggregate with $sample) to read data into the agent's context.
Boundary markers: No specific delimiters or safety instructions are provided to the agent to ensure it ignores malicious instructions embedded within retrieved database content.
Capability inventory: The skill is capable of executing shell commands, performing filesystem writes, and running database queries, including data modification commands (UPDATE, DELETE) with user approval.
Sanitization: The instructions do not specify any sanitization, filtering, or validation of the retrieved database content before processing.
[COMMAND_EXECUTION]: The skill provides numerous shell command templates for database interaction. In reference/mysql.md, the suggested fallback CLI connection method involves passing the database password directly as a command-line argument (-p"$DB4_PASSWORD"). This practice is insecure as command-line arguments can be visible to other users and processes on the same system.
[CREDENTIALS_UNSAFE]: Although the skill correctly advises against printing .env files, the provided command-line snippets demonstrate the usage of raw credentials in shell commands, which can lead to exposure in process lists or shell history buffers.

db-intelligence