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/review-pr/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/review-pr)<a href="https://agentmods.dev/skills/maximhq/bifrost/review-pr"><img src="https://agentmods.dev/badge/skills/maximhq/bifrost/review-pr/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/review-pr"><img src="https://agentmods.dev/badge/skills/maximhq/bifrost/review-pr.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 | $0.00060 | $0.01097 |
| Opus 5.5 | $0.00024 | $0.00439 |
| Sonnet 5.5 | $0.00012 | $0.00219 |
| Haiku 4.5 | $0.00006 | $0.00110 |
Grade A, and why
review-pr 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 2mo 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
PR Review (bifrost custom)
Scope
If a PR number is given: gh pr view <N> --json title,body,author,baseRefName,headRefName,state,additions,deletions,changedFiles and gh pr diff <N> define the scope. Otherwise use git diff @{upstream}...HEAD (fall back to git diff dev...HEAD, then git diff HEAD for uncommitted work).
The diff is the only review scope. When an angle needs surrounding code, Read files in this checkout if it matches the PR branch, otherwise fetch via gh.
Process
- Find - launch independent finder agents in parallel (Agent tool), each returning up to 6 candidates
{file, line, summary, failure_scenario}:- Correctness angles: line-by-line diff scan (read the enclosing function of every hunk); removed-behavior audit (what invariant did deleted/replaced code enforce, and where is it re-established?); cross-file caller/callee trace (grep for changed symbols, check every call site).
- Cleanup angles: reuse (does new code re-implement an existing helper? name it); simplification (redundant state, copy-paste variation, dead code); efficiency (wasted work, repeated I/O, hot-path cost); altitude (is the fix at the right layer, or a bandaid on shared infrastructure?); conventions (check AGENTS.md rules - only flag when you can quote the exact rule and the exact violating line).
- Finders must pass through every candidate with a nameable failure scenario; do not self-censor half-believed ones.
- Verify - dedup candidates that share a mechanism (keep the most concrete scenario), then run one verifier agent per candidate. Each returns exactly one of CONFIRMED (quote the triggering inputs and the wrong output), PLAUSIBLE (mechanism real, trigger uncertain - state what would confirm it), or REFUTED (quote the line that proves it wrong). Drop REFUTED.
Output format (required)
- 2-3 sentence overview of what the PR does and its overall quality.
- Findings table: columns
# | Severity | Location | Finding | Verdict, ranked most-severe first, max 8 rows. Location is a clickablefile:line. If nothing survived verification, say so plainly. - Merge recommendation: exactly one of Merge, Merge after nits, or Not yet - request changes, with a 1-2 sentence reason tied to specific finding numbers. Only CONFIRMED correctness findings may block a merge; PLAUSIBLE and cleanup findings are nits or followups.
- Followups required: numbered list, each tagged "in this PR (blocking)" or "follow-up PR", with concrete file:line targets and a one-line fix description.
- Follow-up PR offer: if any follow-up items exist, offer to prepare them on a new branch. NEVER commit or push - leave changes staged for the user (standing user rule), unless the user explicitly authorizes committing and opening the PR with
gh pr create. - Briefly list refuted candidates with the one-line reason, so the user knows what was checked and cleared.
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.
- 2mo ago First seen · 43 lines · 60 tokens per session scan A a1835dca2168
review-pr is a skill published in the GitHub repository maximhq/bifrost (9,185 stars, last pushed today), licensed Apache-2.0. It adds 60 tokens to every session and 1,097 once invoked, about $0.0002 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
merge-prs
Consume analyzed PR plans, implement authorized fixes with contributor credit, validate exact candidates, and merge only owner-selected and explicitly authorized groups through the shared gates. Requires prior review evidence; unknown or failing required checks remain HOLD.
merge-batch
Prepare and merge an owner-selected PR batch with contributor credit and exact-candidate admission. Use for a pasted PR list; combined focused validation is diagnostic, and every actual landing still requires its applicable checks and authorization.
github
GitHub PR and issue workflows via gh CLI — read-only by default; merge requires approval.
release-pr-review
Review pass on an open release PR (release/ → main) — the step between git-wrapup and release-and-publish when a project releases in gated release PR mode. Reads the PR's commit range through the code-simplifier lens plus a correctness review, verifies whatever an automated reviewer left on the PR, lands fixes as…
pr
Draft a PR body from OMC's paper trail — the smallest visual that proves the change, evidence consumed from the verify protocol (never re-collected), reversibility from the ADR test on the diff, glossary language throughout.
otto-pr-review-loop
Run Polly and OCR on a PR until all findings settle.