The Agent Skills Directory

[COMMAND_EXECUTION]: The audit.py script is vulnerable to SQL injection through the framework parameter in the gaps command. Evidence: The variable is directly interpolated into the SQL string: query += f\" AND c.control_type = '{framework}'\". This allows an attacker to execute arbitrary SQL commands, potentially leading to unauthorized data access.
[REMOTE_CODE_EXECUTION]: The script uses dynamic path manipulation to load shared modules. Evidence: _sys.path.insert(0, str(_Path.home() / \".pi\" / \"skills\")) is used to import TaskClient. Risk: Loading code from computed paths in the home directory is risky as it depends on the security of that specific directory.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. 1. Ingestion points: Data is read from sparta.duckdb. 2. Boundary markers: None present to delimit untrusted data in the output. 3. Capability inventory: Accesses local database files and composes with the create-figure skill. 4. Sanitization: No evidence of escaping or filtering database content before display.

data-audit