Security Policy¶
Supported versions¶
The project is in early alpha (0.x). Only the latest minor version receives
security fixes.
Reporting a vulnerability¶
Please do not open a public issue for security reports.
Send a private email to the maintainer listed on the project's GitHub page with:
- A description of the issue.
- The steps to reproduce.
- The smallest possible test case.
- Your assessment of impact (data exposure, code execution, etc.).
You should expect an acknowledgement within five working days. We will coordinate a fix and disclosure timeline with you.
Threat model¶
This project is an offline CLI that reads local files and writes local files. The 0.1 core has zero network dependencies. Things to keep in mind:
- Untrusted input data. Inventory CSV/JSON files are user-controlled. The toolkit treats them as data, never executes code from them, and never echoes them back as shell input. Reports of injection through input files are in scope.
- Path handling. All file writes happen under
clients/<client>/{data,output,logs}. Reports of path traversal (e.g.--client ../../etc) are in scope. - Future network adapters. When optional adapters (LLM, SERP, Wordstat)
are added, they will read API keys from environment variables or local
.envfiles. The keys must never be logged or written to artifacts. Reports of secret leakage are in scope.
Out of scope¶
- Bugs in third-party services that this toolkit may eventually call.
- Misuse of the tool to scrape sites the user does not own. The CLI itself performs no crawling in 0.1.