Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add neXenio/bifrost-plugin --skill bifrost-code-modegit clone --depth 1 https://github.com/neXenio/bifrost-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/nexenio/bifrost-plugin/bifrost-code-mode)<a href="https://agentmods.dev/skills/nexenio/bifrost-plugin/bifrost-code-mode"><img src="https://agentmods.dev/badge/skills/nexenio/bifrost-plugin/bifrost-code-mode/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/nexenio/bifrost-plugin/bifrost-code-mode"><img src="https://agentmods.dev/badge/skills/nexenio/bifrost-plugin/bifrost-code-mode.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00146 | $0.01798 |
| Opus 5.5 | $0.00058 | $0.00719 |
| Sonnet 5.5 | $0.00029 | $0.00360 |
| Haiku 4.5 | $0.00015 | $0.00180 |
Grade A, and why
bifrost-code-mode scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 178 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Bifrost code mode
Most servers on a Bifrost gateway are not flat tools. They do not appear in your
tool list and <prefix><server>-<tool> does not exist for them (mcp__plugin_bifrost-plugin_bifrost__ from the plugin, mcp__claude_ai_luca_Bifrost__ from the claude.ai connector, mcp__bifrost__ from a hand-added server). They are reached
by writing a short Starlark snippet and running it with executeToolCode.
If a capability seems missing, it is almost certainly here. Do not conclude the gateway lacks it, and do not fall back to guessing from training data, until you have checked.
The four meta-tools
Exact schemas, as the gateway advertises them:
| Tool | Parameters | Required |
|---|---|---|
listToolFiles |
(none) | — |
readToolFile |
fileName, startLine, endLine |
fileName |
getToolDocs |
server, tool |
both |
executeToolCode |
code |
code |
The parameter is fileName, not path or file. Passing path fails with
fileName parameter is required and must be a string.
fileName is the catalog path exactly as listToolFiles() prints it. On our gateway
that is one file per server, servers/<server>.pyi; gateways with tool-level binding
use servers/<server>/<tool>.pyi instead.
Workflow
1. Discover. listToolFiles() returns the catalog, one file per server:
servers/
gitlab.pyi
jira.pyi
...
2. Confirm the signature. Never call a tool whose parameters you have not read.
readToolFile(fileName="servers/jira.pyi")
returns typed stubs for every tool on that server, with the exact callable names and inline descriptions:
# jira server tools
# Usage: jira.tool_name(param=value)
def search_jira_tickets(query: str, max_results: float = None, service: str = None, status: str = None) -> dict: # Search JIRA before filing a new ticket to avoid duplicates.
The def name is the exact callable. Descriptions in .pyi files may be truncated —
use getToolDocs(server="jira", tool="search_jira_tickets") when a parameter's meaning
or return shape is unclear.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago Changed · +26 lines 644878e2e95a
- 16d ago Changed · +34 tokens per session 8b910424e9cb
- 1mo ago First seen · 152 lines · 112 tokens per session scan A 1e4f9e20b46e
bifrost-code-mode is a skill published in the GitHub repository neXenio/bifrost-plugin (4 stars, last pushed 2d ago), licensed MIT. It adds 146 tokens to every session and 1,798 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
oatda-list-models
Use when the user wants to list available AI models from OATDA's 10+ providers. Filter by type (chat, image, video, audio) or by provider name, including speech, transcription, and translation models.
oatda-compare-models
Use when the user wants to compare outputs from multiple LLM models side-by-side in a single request. Sends one prompt to multiple providers (OpenAI, Anthropic, Google, DeepSeek, etc.) in parallel via OATDA's /api/v1/compare endpoint and returns each model's response. Triggers on "compare models", "which model is…
oatda-generate-image
Use when the user wants to generate images using AI models through OATDA's unified API. Supports DALL-E 3, GPT-Image-1, Google Imagen, MiniMax, Qwen, and xAI image models.
oatda-generate-speech
Use when the user wants to generate speech/audio from text using OATDA's unified audio API. MCP generatespeech returns a short-lived signed AUDIOURL (download with curl, no login). HTTP /api/v1/llm/speech returns a file via curl --output. Supports OpenAI TTS, xAI grok-tts, voiceovers, and accessibility audio.
oatda-generate-video
Use when the user wants to generate videos using AI models through OATDA's unified API. Supports MiniMax, Google Veo, Alibaba Wan, ZAI, and OpenAI Sora. Video generation is asynchronous.
oatda-transcribe-audio
Use when the user wants to transcribe audio to text using OATDA's unified audio API. Supports speech-to-text (STT), meetings, podcasts, voice notes, Whisper-style transcription, and the transcribeaudio MCP capability.