Skip to content

Security: fir3storm/bramhashiv-ai

Security

SECURITY.md

Security & Credentials

Where your secrets actually live

BramhaShiv does not bundle, copy, or commit any credentials. All API keys, OAuth tokens, and session data stay on your machine in OpenCode's user-private locations:

~/.local/share/opencode/auth.json   — OAuth + API keys (provider auth)
~/.config/bramhashiv/catalog.yaml   — your edited model catalog
~/.config/bramhashiv/state.json     — pinned model + last classifier output
~/.config/bramhashiv/overrides.log  — local-only telemetry of override events

These paths are outside the git repository and never tracked. Cloning this repo gives you zero access to anyone else's credentials.

What you need to provide yourself

When you install BramhaShiv, you must supply your own provider auth via OpenCode:

opencode providers login -p openai        # ChatGPT OAuth (uses your subscription)
opencode providers login -p google        # paste your own Gemini API key
opencode providers login -p huggingface   # paste your own HF token
opencode providers login -p openrouter    # paste your own OpenRouter key
opencode providers login -p anthropic     # paste your own Anthropic API key (optional)

The plugin reads from ~/.local/share/opencode/auth.json at activation. If a provider isn't authed, BramhaShiv automatically filters its models out of routing — you'll never accidentally try to dispatch through a missing credential.

What's in this repo

  • TypeScript source code (src/, tests/)
  • A model scoring catalog (config/default-catalog.yaml) — opinions about model strengths, no credentials
  • Build / test config (package.json, tsconfig.json, bunfig.toml)
  • Documentation

Test files contain string literals like "AIza-test-key" and "hf_x". These are intentionally invalid placeholder strings used as fixtures. They are not real credentials and cannot authenticate against any service.

Verifying for yourself

Run a credential scan against a fresh clone:

git clone https://github.com/fir3storm/bramhashiv-ai.git
cd bramhashiv-ai

# Live working tree
git grep -nE 'AIza[A-Za-z0-9_-]{15,}|hf_[A-Za-z0-9]{20,}|sk-[A-Za-z0-9]{20,}|gh[op]_[A-Za-z0-9]{30,}|eyJ[A-Za-z0-9._-]{50,}'

# Full git history
git log --all -p | grep -E 'AIza[A-Za-z0-9_-]{15,}|hf_[A-Za-z0-9]{20,}|sk-[A-Za-z0-9]{20,}'

Both should return nothing. If you find anything that looks like a real credential, please file an issue immediately.

Reporting security issues

If you discover a vulnerability or accidental credential leak, please:

  1. Do not open a public issue.
  2. Email the maintainer directly (see package.json author field) with the details and a minimal reproduction.

Threat model recap

  • What this plugin does: route your text-completion requests across providers you've already authenticated.
  • What this plugin does NOT do: capture, transmit, log, or share any of your credentials with anyone.
  • What you control: every API key and OAuth token belongs to you, lives in your home directory, and never leaves your machine.

There aren’t any published security advisories