sqlite-inspector

Warn

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on docker exec to run sqlite3 queries inside MikoPBX Docker containers. This gives the agent the ability to execute commands within the containerized environment.
  • Evidence in SKILL.md: The Quick Start section provides commands like docker exec <container_id> sqlite3 /cf/conf/mikopbx.db "SELECT * FROM m_Extensions LIMIT 5".
  • Evidence in scripts/db_query.sh: The core logic of the helper script uses docker exec "$CONTAINER_ID" sqlite3 "$DB_PATH" "$query" $format_args to run queries.
  • [CREDENTIALS_UNSAFE]: The skill provides documentation and specific SQL query patterns for accessing highly sensitive configuration data, including plaintext secrets and administrative credentials.
  • Evidence in reference/schema-reference.md: Lists sensitive table columns such as m_Sip.secret (SIP password), m_PbxSettings.SSHPassword, and m_PbxSettings.WebAdminPassword.
  • Evidence in reference/common-queries.md: Provides ready-to-use queries like SELECT s.extension, s.secret FROM m_Sip s and SELECT * FROM m_PbxSettings WHERE key IN ('SSHPassword', 'WebAdminLogin', 'WebAdminPassword').
  • [PRIVILEGE_ESCALATION]: The documentation instructs the agent on how to acquire higher privileges to bypass database lock or permission errors by running commands as the root user within the container.
  • Evidence in SKILL.md: The Troubleshooting section recommends docker exec -u root <container_id> sqlite3 ... if permission is denied.
  • [INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface where an agent might be manipulated into running malicious SQL queries or exposing sensitive data if it processes untrusted user input without sanitization.
  • Ingestion points: Untrusted data enters the context through the query argument in scripts/db_query.sh or through the interactive docker exec commands in SKILL.md.
  • Boundary markers: None identified; there are no instructions for the agent to sanitize or validate external inputs before using them in database queries.
  • Capability inventory: The skill provides full read access to the database via docker exec, including access to authentication settings and call history.
  • Sanitization: None identified; the script passes the query string directly to the database engine.
  • [DYNAMIC_EXECUTION]: The helper script scripts/db_query.sh dynamically constructs shell commands by interpolating arguments into a command template.
  • Evidence in scripts/db_query.sh: The function execute_query builds a command string using $CONTAINER_ID, $DB_PATH, and $query variables which are then executed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 2, 2026, 06:27 AM
Security Audit — agent-trust-hub — sqlite-inspector