Bifrost is an AI gateway that gives applications one OpenAI-compatible API for accessing models from more than 23 providers. It is used to route model requests, handle provider failover and load balancing, and apply features such as caching, guardrails, and MCP gateway support.
Borrowing it
Nothing to install: this file belongs to maximhq/bifrost. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/maximhq/bifrost/dev/.claude/skills/harness-test-writer/SKILL.mdgit clone --depth 1 https://github.com/maximhq/bifrostWrote 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/maximhq/bifrost/harness-test-writer)<a href="https://agentmods.dev/skills/maximhq/bifrost/harness-test-writer"><img src="https://agentmods.dev/badge/skills/maximhq/bifrost/harness-test-writer/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/maximhq/bifrost/harness-test-writer"><img src="https://agentmods.dev/badge/skills/maximhq/bifrost/harness-test-writer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00133 | $0.02478 |
| Opus 5.5 | $0.00053 | $0.00991 |
| Sonnet 5.5 | $0.00027 | $0.00496 |
| Haiku 4.5 | $0.00013 | $0.00248 |
Grade A, and why
harness-test-writer 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 1mo 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 — 184 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Provider Harness Test Writer
Turn a PR or GitHub issue into regression coverage in the provider harness: the Postman
collection at tests/e2e/api/collections/provider-harness.json, executed by newman via
make run-provider-harness-test.
This is NOT the Go test harness in core/internal/llmtests/ (that one is run via
make test-core PROVIDER=...). If the user seems to want Go-level scenario tests,
confirm before proceeding.
Step 1 - Resolve the reference
The argument may be a PR number, an issue number, or a GitHub URL.
- URL containing
/pull/=> PR. URL containing/issues/=> issue. - A bare number is ambiguous: try
gh pr view <N> --repo maximhq/bifrostfirst; if it 404s, trygh issue view <N>. If BOTH exist and refer to different things, ask the user which one they mean with AskUserQuestion. - No argument at all => ask the user for the PR/issue reference.
Fetch full context:
gh pr view <N> --repo maximhq/bifrost --json title,body,state,files,baseRefName
gh pr diff <N> --repo maximhq/bifrost
# or
gh issue view <N> --repo maximhq/bifrost --json title,body,state,labels,comments
For a PR, also note any Closes #X issue and fetch that issue too - the issue usually
contains the client-visible reproduction (exact request shapes, error bodies, status
codes) that the harness case must mirror.
Step 2 - Understand the wire-level behavior to pin
The harness tests Bifrost from the outside: HTTP requests against gateway routes, with assertions on status codes and response/SSE bodies. Translate the PR/issue into that frame:
- Which route(s)? (
/v1/chat/completions,/v1/responses, drop-ins like/openai,/anthropic,/bedrock,/genai, passthrough routes, ...) - Which provider(s) and model(s)?
- What request shape triggers the bug/feature? Reconstruct it from the issue repro or from the code path in the diff (read the changed functions and their callers).
- What is the observable failure signature before the fix (status code, error body substring) and the expected behavior after?
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.
- 1mo ago Changed · +8 lines b4a29fa8e3b0
- 1mo ago Changed · +1 lines 4dd7113d02fc
- 2mo ago First seen · 175 lines · 133 tokens per session scan A 04b488900d23
harness-test-writer is a skill published in the GitHub repository maximhq/bifrost (9,185 stars, last pushed today), licensed Apache-2.0. It adds 133 tokens to every session and 2,478 once invoked, about $0.0005 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 2026-08-30.
Other skills, from other repositories
Browser Agent QA Testing
Teach agents to use AI browser agents for exploratory and smoke QA with step budgets, evidence-based assertions, guardrails, and Playwright conversion.
verify
A local, evidence-based procedure for checking that an implemented feature or bug fix works in the actual app. It uses a browser, test accounts, logs, and screenshots or command output rather than relying only on linting or type checks.
test-sweep
Run all test suites (unit, integration, API, E2E) and aggregate results into a summary report. Use after completing execution slices or before the Review Gate.
audit-loop
Run a recursive audit drain loop — discover findings from the running system, triage each into bug/spec/classifier lanes, repeat until convergence. USE FOR: end-to-end audit of a deployed or locally-running app, draining findings to zero. DO NOT USE FOR: single-shot tempering runs (use forgetemperingrun), one-off bug…
qa-e2e-security
Real-cluster security regression QA — backend auth enforcement, grant/session policy, gateway deny paths, audit emission, trust escalation, UI/API security headers, HTTPS redirect modes, and secret-leak scanning in live logs. Use when Codex is asked to verify a change did not regress auth, governance, gateway policy…
qa-e2e-operations
Real-cluster operational QA for MCP Runtime — operator/CRD reconciliation, CLI flows, setup/test-mode regressions, registry pulls, ingress wiring, rollout health, and observability — against the live Kind contributor cluster. Use when Codex is asked to verify a change does not regress operator, CLI, setup, registry…