okf-postgresql

Pass

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The runIngest function in main.go is vulnerable to potential SQL injection through improper handling of database identifiers.
  • Evidence: Lines 301 and 331 in main.go use fmt.Sprintf to construct COMMENT ON statements using schema, table, and column names directly from the input bundle without identifier quoting: fmt.Sprintf("COMMENT ON TABLE %s.%s IS '%s'", ...).
  • Risk: Although the skill attempts to verify object existence prior to execution, the lack of proper identifier quoting (which is correctly implemented elsewhere in the skill using quoteIdent) allows for malicious SQL manipulation if the ingested bundle content is untrusted.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data that could influence its operation on sensitive systems.
  • Ingestion points: The ingest command parses markdown files within an OKF bundle from the local file system (e.g., ReadConceptDoc in main.go).
  • Boundary markers: No boundary markers or "ignore" instructions are used when processing the table and column descriptions from markdown.
  • Capability inventory: The skill possesses the capability to execute SQL commands on a PostgreSQL database and write to the local file system.
  • Sanitization: While single quotes in comment values are escaped, the structural elements (identifiers) are not sanitized or quoted.
  • [EXTERNAL_DOWNLOADS]: The skill downloads and builds dependencies from external repositories.
  • Fetches the okf-go core library from the author's GitHub repository (github.com/xSAVIKx/okf-skills/okf-go).
  • Utilizes the github.com/lib/pq PostgreSQL driver.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 12, 2026, 08:04 AM
Security Audit — agent-trust-hub — okf-postgresql