okf-graphql
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted GraphQL SDL files to generate documentation, creating an attack surface where malicious instructions in schema descriptions could influence downstream agents.\n
- Ingestion points: GraphQL SDL documents provided via the
--schemaCLI flag inmain.go.\n - Boundary markers: The generated Markdown output lacks explicit delimiters or instructions for agents to ignore embedded text.\n
- Capability inventory: The skill performs filesystem operations, including directory creation (
os.MkdirAll) and file writing (okf.WriteConceptDoc), as seen inmain.go.\n - Sanitization: Employs
okf.SanitizeCelland custom slugification to format output, though these do not filter for natural language instructions.\n- [EXTERNAL_DOWNLOADS]: Fetches and installs the connector binary from the author's repository.\n - Evidence: Setup instructions in
SKILL.mdusego install github.com/xSAVIKx/okf-skills/skills/okf-graphql@latest.
Audit Metadata