review-and-provider-backends
Configure and troubleshoot ARIS cross-model reviewers, MCP bridges, OpenAI-compatible providers, MiniMax, Claude/Gemini overlays, manual review, and optional Feishu/Lark integrations.
- License 许可
- MIT
- 最近更新
- 2026-09-03
- Stars
- ★ 351
- Forks
- 27
01 前言 Front Matter
- name
- review-and-provider-backends
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Configure and troubleshoot ARIS cross-model reviewers, MCP bridges, OpenAI-compatible providers, MiniMax, Claude/Gemini overlays, manual review, and optional Feishu/Lark integrations.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/mcp-servers.md2.1 KB
- 参考文档references/provider-matrix.md2.4 KB
- 参考文档references/troubleshooting.md1.7 KB
03 SKILL.md 正文
Review and Provider Backends
Use this sub-skill when ARIS needs an external reviewer, MCP server registration, an alternative LLM endpoint, a manual human review handoff, a Claude/Gemini/Codex combination, MiniMax or ModelScope configuration, Feishu/Lark notification, or a Codex image bridge.
Route Here
- Configure the default Claude executor + Codex MCP reviewer path.
- Configure Codex executor + Claude/Gemini reviewer overlays.
- Use a generic OpenAI-compatible
llm-chatbackend or MiniMax instead of Codex MCP. - Run manual review in browser or headless file mode.
- Diagnose provider credentials, model names, retries, timeouts, MCP registration, protocol negotiation, and restart requirements.
- Preserve cross-model independence and provenance evidence.
Reroute
- Host skill directory, manifest, or selective install:
../install-and-distribution/SKILL.md. - Workflow selection and assurance levels:
../workflow-routing-and-skill-catalog/SKILL.md. - Research state, trace files, or watchdog:
../state-recovery-and-experiment-ops/SKILL.md. - Source edits or mocked server tests:
../repository-maintenance/SKILL.md.
Safe Backend Pattern
- Identify executor family and reviewer family.
- Pick one reviewer route and record its model, server name, endpoint, and credential source.
- Register MCP before invoking a workflow, then restart the host agent.
- Validate with a trivial tool call before spending a full review request.
- Keep raw review artifacts and traces; never let the executor summarize away a failure.
- If a provider is unavailable, switch to an explicitly configured alternative or mark the review blocked.
Reference Map
references/provider-matrix.mdcompares reviewer/provider routes, required environment variables, and independence implications.references/mcp-servers.mdsummarizes the optional server contracts and testable failure modes.references/troubleshooting.mdcovers keys, endpoints, protocol, model, timeout, and local handoff failures.- Root
../../references/repo-provenance.mdand../../references/troubleshooting.mdprovide provenance and cross-cutting diagnosis.
Avoid
- Never publish API keys, bearer tokens, raw provider error bodies, or local debug paths in artifacts.
- Do not call a same-family reviewer and label it independent.
- Do not treat mocked MCP unit tests as proof of live provider connectivity.
- Do not install every provider dependency when one selected reviewer path is enough.
04 同仓库更多 Skills
passnet-triton-opt
Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.
passnet-solve
END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.
passnet-skill
PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.
passnet-pattern-fusion
Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.
passnet-orchestrate
Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.
passnet-feedback
Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).