okf-openapi
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPERSISTENCEEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from OpenAPI specifications via the --spec parameter. * Ingestion points: The runProduce and runIngest functions in main.go read specification files. * Boundary markers: The generated Markdown documentation lacks delimiters or warnings to ignore embedded instructions in the spec's description fields. * Capability inventory: The tool has file-write capabilities (os.WriteFile, os.MkdirAll) used to create the documentation bundle. * Sanitization: There is no sanitization of the input metadata to prevent instructions from being interpreted by an AI agent reading the output.
- [PERSISTENCE]: The produce command in main.go generates concept files using schema names directly from the specification. The filepath.Join operation does not validate against path traversal sequences (e.g., '../') in schema names. While a .md extension is always appended, this could potentially allow writing files to unintended directories if a crafted specification is used to target sensitive locations.
- [EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md recommend installing the binary using go install github.com/xSAVIKx/okf-skills/skills/okf-openapi@latest. This fetches the source code from the author's public repository for compilation and installation.
Audit Metadata