在本地磁盘以分类纯 Markdown 文件保存需要长期留存的事实,与智能体内置记忆并存。
编程
Open Feed Recsys Reviewer
试用Review open feed recommendation repositories with source-backed evidence, artifact-readiness checks, and cautious architecture summaries. Use when a user ask...
它能做什么
Use this skill to review public feed recommendation repositories and separate what the source supports from what remains unproven.
技能文档
Open Feed Recsys Lab
Use this skill to review public feed recommendation repositories and separate what the source supports from what remains unproven.
Default target: https://github.com/xai-org/x-algorithm.
Review Workflow
- Confirm the target repository, commit, public link, or user-provided file set.
- Build an evidence ledger with repository/ref, files inspected, supported claims, weak claims, and open questions.
- Separate source inspection from runnable-model proof. If Phoenix artifacts are missing, say execution is not verified.
- Map the visible architecture: Home Mixer, Phoenix retrieval/ranking, Thunder, Grox, ads blending, and candidate pipelines when present.
- Treat public algorithm claims as hypotheses. Mark each claim
supported,partly_supported,unsupported, ornot_public_repo. - Frame shareable output around reproducibility and evidence. Do not present the public source as a reach predictor or full live-platform clone.
Phoenix Artifact Readiness
Only treat Phoenix execution as verified after the official artifacts are present and the user has supplied a local run result. The expected extracted directory is:
phoenix/artifacts/oss-phoenix-artifacts/
Expected files include:
retrieval/model_params.npzretrieval/embedding_tables.npzretrieval/config.jsonranker/model_params.npzranker/embedding_tables.npzranker/config.jsonsports_corpus.npzexample_sequence.json
If the report says Phoenix is not run-ready, state that source inspection succeeded but execution is blocked by missing extracted LFS artifacts.
Boundaries
- Use public source or user-provided local files only.
- Do not request or inspect private user data, authenticated account pages, or non-public analytics.
- Do not promise reach, ranking, virality, revenue, or live-platform equivalence.
- Do not claim that public source contains every production weight, threshold, model version, or serving rule.
- Do not execute repository code or delete local files as part of a review unless the user separately asks for that engineering work.
Output Shape
Return a concise review with:
Target: repo, ref, and source date when known.Supported: claims backed by public source.Blocked: missing artifacts, missing configs, or live-platform gaps.Architecture: short component map.Risks: overclaims, ambiguous evidence, or unsafe product positioning.Next check: the smallest concrete verification step.
The preferred public-safe framing is:
I verified the open X For You algorithm locally at commit ; here is the report.
Companion Skills
Use x-algo-claim-auditor when the user asks whether a public claim, screenshot text, or viral thread is supported by the source.
Use tinytroupe-feed-research-lab when the user asks to compare draft posts, simulate audience reactions, or pretest a post before publishing. Keep that work labeled as synthetic audience research, not source verification or reach prediction.
相关技能
把自然语言描述转为结构化 JSON,并由 mcp-diagram-generator MCP 服务生成 Draw.io、Mermaid 或 Excalidraw 图表文件。
从 AdMapix API 拉取广告创意、应用、榜单和收入预估等数据,原样返回结构化 JSON。
以 AI 机器人身份加入视频会议,提供语音、虚拟形象与屏幕共享四种模式。
诊断生产力系统反复失效的根因,给出最小干预——容量测算、瓶颈定位、可靠的本地记录。
编写、调试与调优 Playwright 测试,涵盖定位器策略、追踪诊断与 CI 友好的超时配置。
zack-dev-cm 的更多技能
浏览全部技能OpenClaw CWS Publisher is a public ClawHub Chrome Web Store publisher skill. Use it when the user says "chrome web store publisher", "extension release publi...
SOTA Agent is a public ClawHub SOTA-campaign skill for CV and DS work. Use it when the user says "sota agent", "state of the art benchmark scouting", or want...
Use when studying or normalizing Chinese lesson transcript/subtitle inputs with bundled public lesson data, learner docs, local export bundles, and pilot-fir...
OpenClaw YouTube Publisher is a public ClawHub YouTube publishing skill. Use it when the user says "youtube creator ops", "youtube publisher", "YouTube Studi...
Build source-backed research claim ledgers from drafts, literature matrices, notes, citation lists, PDFs, source packets, or reviewer comments. Use when the...
Install, wire, audit, and harden AgentMemory for AI coding agents and project workspaces. Use when Codex needs to adopt rohitg00/agentmemory, configure Codex...