Email anmolnoor59@gmail.com with a clear description and minimal reproduction. Do not open a public issue for security-relevant bugs.
I'll acknowledge the report on a best-effort basis (this is a personal project, not a service). Please give me a reasonable window to investigate and patch before public disclosure.
Only the latest release of foundation-cli on main receives security fixes. There are no LTS branches.
In scope:
- The CLI itself (
src/foundation/) - The capability execution surface (file/git operations, shell capability)
- The history database and event log
- Any configuration parsing
Out of scope:
- Vulnerabilities in upstream dependencies (please report those upstream — I'll bump versions when patches land)
- Issues that require already-compromised local credentials or filesystem access
- Bugs in third-party LLM providers (OpenAI, Ollama, etc.)
Foundation CLI executes shell commands and writes files on your machine. Running it against an untrusted prompt or model output is, by design, equivalent to giving a stranger your shell. The approval gate is the primary defense. If you find a way to bypass approvals or trigger side effects without an explicit [approve], that's a vulnerability — please report it.