Claude Skills by NeverSight

Sync experiment results from the code repo into the paper's daily experiments log (daily_experiments.tex). Use when you have new experiment results to record, want to update the paper with latest numbers, or log experimental findings from an ML research project.

Plan and write strategic rebuttals after real paper reviews arrive. Use this skill whenever the user has OpenReview reviews, reviewer comments, scores, confidence ratings, meta-reviews, author response windows, or wants to decide which experiments to run, infer reviewer intent, draft point-by-point responses, prepare follow-up discussion replies, or improve wording after reviews for ML/AI venues such as NeurIPS, ICML, ICLR, CVPR, ACL, EMNLP, or similar conferences.

Produce a multi-paper comparison matrix across a literature corpus with tiered read depth. Use when multiple papers need to be compared side-by-side for method differences, performance gaps, closest-work ranking, or trend identification — distinct from per-paper source cards (reference-reading-summarizer) and single-paper project linking (reference-project-synthesizer).

Manage project reference sources under reference/. Use when scanning, ingesting, indexing, deduplicating, monitoring, or tracking processing status for papers, PDFs, Word docs, Markdown notes, BibTeX files, scripts, specs, or source bundles without deeply reading them.

Connect structured reference source cards to the active ML project. Use when papers, collaborator docs, Markdown notes, specs, scripts, BibTeX files, or source bundles should inform claims, risks, baselines, benchmarks, experiments, algorithm design, implementation, writing contracts, citations, collaborator actions, project initialization, or memory writeback.

Read and summarize project reference sources into structured source cards. Use for skimming papers, PDFs, Word docs, Markdown notes, BibTeX files, scripts, specs, collaborator feedback, or source bundles; extract writing patterns, methods, theory, benchmarks, baselines, implementation hints, risks, constraints, and project seeds without yet deciding project implications.

Draft ML/AI related work as novelty-boundary writing. Use for closest-work grouping, citation roles, paragraph plans, boundary statements, and safe novelty wording.

Prepare and publish a research code repository for public release alongside a paper (arXiv, conference, GitHub). Use when the user wants to open-source code, create a GitHub release, package a code submission, make code public, or prepare a reproducibility release.

Coordinate local, Git remote, and SSH/HPC/RunAI research projects. Use for server state, sync safety, job submission, interactive sessions, logs, artifact lookup, context recovery, raw SSH commands, remote shell one-liners, SSH quoting issues, remote-cmd, remote-bash, or avoiding local shell expansion of remote variables.

Help a CS or AI PhD student turn a rough research idea into a validated next-step decision using the FIVE+C framework. Use this skill whenever the user says they have a research idea, wants to know whether an idea is worth pursuing, needs help choosing between project directions, is preparing to pitch an idea to an advisor or senior student, or feels unsure whether a project is too incremental, too ambitious, already solved, hard to evaluate, or missing resources.

Initialize, inspect, and maintain a hierarchical memory system for an ML research project across paper, code, worktrees, slides, reviewer simulation, rebuttal, experiments, claims, evidence, risks, and actions. Use this skill whenever the user wants cross-session project memory, project bootstrapping context, feedback-loop tracking, claim-evidence-risk-action alignment, worktree memory, or consistency between code results, paper writing, slides, reviews, and rebuttal.

Use when auditing completed results for confounds, claim-drift, protocol integrity, or attribution before locking claims into the paper. Not for deciding what to do after a surprising result (use result-diagnosis). Not for significance tests or effect sizes (use statistical-analysis-planner). Not for engineering failures (use experiment-debugger).

Design and write reusable research slide decks. Use for advisor updates, lab talks, reading reports, proposals, conference talks, Slidev content, and slide structure.

Diagnose surprising or negative ML/AI results. Use when methods fail, metrics conflict, seeds vary, baselines win, plots look suspicious, or next action is unclear.

Submit or run an ML experiment on a compute environment (local, SLURM HPC, RunAI/Kubernetes). Use when the user wants to launch a training run, submit a job, run ablations, or execute an experiment script on any compute cluster.

Use when probing the status of an existing job — queued, stuck, running, or finished — across local, SLURM, RunAI, or SSH. Not for launching new jobs (use run-experiment). Not for debugging NaN/OOM/engineering failures (use experiment-debugger). Not for interpreting valid but surprising results (use result-diagnosis).

Perform common Git operations safely with sandbox-aware failure handling. Use whenever the user wants to inspect or modify git state, especially for cherry-pick, merge, rebase, commit, branch, stash, or worktree workflows. Always use this skill when the user mentions a Git failure, conflict, cherry-pick, merge issue, worktree, branch checkout problem, lock file, permission denied, operation not permitted, or any case where a sandboxed agent might confuse an environment restriction with a real...

Run artifact-driven sidecar agent tasks through one-shot Codex CLI sessions. Use when a main agent should delegate bounded scans, drafts, audits, pre-reviews, or mechanical repo tasks to a fast isolated sidecar model such as gpt-5.3-codex-spark while keeping final decisions with the main agent.

Audit a skill repository or installed skill collection for global consistency, lifecycle coverage, routing quality, documentation drift, memory writeback coverage, stale future-skill references, broken helper paths, and validation readiness. Use this skill whenever the user asks for a global consistency audit, skill taxonomy review, lifecycle audit, cross-skill routing audit, README or AGENTS inventory consistency check, or maintenance pass over a collection of agent skills.

Plan and report statistical rigor for ML experiment results. Use when significance testing, effect size reporting, confidence intervals, seed variance analysis, or multiple-comparison corrections are needed before including results in a paper or rebuttal.

Pre-submission checklist for LaTeX academic papers. Use when the user wants to submit a paper, check submission readiness, prepare camera-ready, switch to final mode, or verify a paper is ready for a conference deadline.

Review ML/AI result tables, LaTeX table files, captions, provenance, and paper table style. Use for benchmark, ablation, metric, model-spec, and compute tables.

Audit project token usage from local Codex and Claude Code logs. Use when the user asks to measure token burn, token consumption, project attention, agent usage, Codex/Claude Code usage, token efficiency, or lifecycle telemetry for a project.

Analyse code changes since the last docs update and refresh the project's documentation files. Use when code has changed and documentation needs to be updated, after implementing new features, or before a milestone commit.

Build a retrospective or forward-looking work timeline from git commits, project docs, user notes, or chat records, then output a Markdown and/or HTML report with a Gantt chart or timeline visualization. Use when the user wants to review past work across one or more projects, explain time allocation to a mentor, summarize what was done in a period, or plan the next phase with a timeline.

Use when results are valid but surprising, negative, unstable, or ambiguous — to decide debug/rerun/ablate/revise/park. Not for engineering failures like NaN/OOM (use experiment-debugger). Not for confound or claim-drift audit before locking results into the paper (use research-results-auditor).

Flaskp proxy server from Cerebase to OpenClaw

將開源專案的 PO/POT 翻譯並驗證成正體中文（zh-TW）PO 檔的 SOP 與工具。當使用者提到翻譯、在地化、l10n、zh_TW／zh-Hant、正體中文、gettext、PO／POT 檔或 Weblate 上傳時使用。

Reviews Java 25 and Spring Boot 4 codebases, pull requests, files, and modules for migration risks, architecture boundary violations, JSpecify null-safety issues, security flaws, performance regressions, and Spring Data pitfalls. Use when the task is a concrete Java or Spring code review with code context. Do not use for Kotlin-only code, non-Spring frameworks, or generic review advice without files or diffs.

Comprehensive code review for Java 25 and Spring Boot 4 apps. Use when reviewing, checking, auditing, or analyzing Java/Spring Boot code (files, modules, PRs, or full codebases) for migration risks, Spring Boot 4 best practices, JSpecify null-safety, security vulnerabilities, performance bottlenecks, data access pitfalls, architecture boundaries (DDD/Hexagonal/Spring Modulith), or modern Java 25 usage.

Implement Spring Data JPA repositories, entities, and queries following modern best practices. Use for creating repositories (only for aggregate roots), writing queries (@Query, DTO projections), custom repositories (Criteria API, bulk ops), CQRS query services, entity relationships, and performance optimization. Covers patterns from simple repositories to advanced CQRS with detailed anti-patterns guidance.

Migrate Spring Boot projects to version 4.0 with Java 25, including Spring Modulith 2.0 and Testcontainers 2.x upgrades. Use when user requests upgrading Spring Boot, migrating to Java 25, updating dependencies to Spring Boot 4, mentions Jackson 3 migration, asks about starter renames (web→webmvc, aop→aspectj), fixing test annotations (@MockBean→@MockitoBean), or needs help with Spring Modulith 2.0 or Testcontainers 2.x compatibility. Analyzes codebase for migration issues and guides through ...

Systematic naming for deployment nodes (VMs, zones, environments) enables consistent infrastructure documentation. Use {Environment}{Service}Vm for VMs, {Tier} for zones, and include VLAN/network details.

Surface concerns, questions, and needs from each stakeholder perspective (users, photographers, facilities). Deep-dives into anxieties and expectations.

Factory skill that creates production-grade, benchmarked, autonomously improved, and verified agent skills. Orchestrates a 5-phase pipeline: interview the user to discover purpose and gold standards, research domain materials with parallel subagents, draft the skill with a design-first approach, invoke autoresearch to iterate against gold-standard-driven LLM-as-judge evaluation, and verify quality through multi-agent consensus with a devil's advocate. Use when building a new skill, creating a...

Principle-engineering posture for production-grade code. Problem-classification (A/B/C) before implementing, plans before code, simplest-correct-solution-first, math-first (closed-form before loop), ACM-grade algorithms, EXPLAIN-first databases with separate data-migration plans, never-N+1, idempotent-atomic writes, realtime-first, concurrent-by-default, graph-aware schemas, recovery-over-crash, validate-at-borders-assert-internally, observable-surfaces-as-architecture (SEO/a11y/perf), AI/LLM...

Runs a doer -> verifier-panel -> consensus loop to verify a deliverable before it ships. An orchestrator freezes acceptance criteria before implementation, dispatches a doer, then convenes a context-walled panel of independent verifiers - including an adversary with an explicit must-oppose mandate - for evidence-anchored review adjudicated to a SHIP / SHIP_WITH_CAVEATS / ITERATE / BLOCK / ESCALATE verdict logged to a ledger. Use for multi-agent verification of any artifact - code slices, plan...

依據 git diff 自動產生 Conventional Commits 格式的 commit message 與完整 git 指令。 輸入「cm」時啟用。

Convert LaTeX text in Word .docx files into MathType or native Word equation objects on Windows. Use when working with docx formula conversion, turning standalone or inline LaTeX into equations, cleaning stray display-$ characters, or safely handling Chinese \\text{} content. 在 Windows 上将 Word .docx 中的 LaTeX 文本转换为 MathType 或 Word 公式对象，用于 docx 公式转换、显示公式残留 $ 清理以及中文 \\text{} 安全处理。

Author a review-ready greenfield `blueprint.md` contract for a new or baseline-only Python repository, reusing the locked blueprint v1 schema already consumed by `python-project-init-greenfield`.

Coding style guidelines for the `results` library. Use this when writing, reviewing, or modifying any Python code that uses Result, Maybe, Ok, Err, Some, or Nothing types.

Flussonic Media Server expert — configuration, API, live streaming, transcoding, DVR, restreaming, CDN, protocols, and troubleshooting. Use this skill whenever the user mentions Flussonic, media server streaming configuration, live ingest setup (RTMP/SRT/WebRTC/multicast sources), transcoding with hardware acceleration (NVENC/QSV), DVR recording and playback, restreaming to CDN or social media, HLS/DASH/LL-HLS/WebRTC delivery, DRM protection, IPTV/OTT setup, Flussonic API endpoints, cluster/l...

Wrap a Rust function in a Jolt zero-knowledge proof

Structured 19-area integrity audit for subtle bugs, design flaws and operational correctness risks. Run after incidents or refactors, before merging large branches or as a periodic health check.

Structured discovery process for starting new work. Aligns on goals, success criteria, and key design decisions through a two-phase interview: goal clarification then systematic resolution of design questions with options, tradeoffs and recommendations. Use when kicking off a new project or feature.

Turns a Brief into a Map using domain-driven design: ubiquitous language, subdomains, bounded contexts, context map, key domain flows, per-context design and architecture decisions. Use after 'brief' and before 'plan'. Use when the user has a completed Brief and wants to map out the domain.

Turns a Map into a Plan using dependency-aware thin vertical implementation slices. Use after 'map' and before 'raise'. Use when the user has a completed Map and wants to break it into ordered slices.

Turns a Plan into dependency-aware beads issues. Use after 'plan' and before 'ship'. Use when the user has a completed Plan and wants to create trackable issues.

Implement beads issues using test-driven development. Use after 'raise'. Use when the user has open beads issues and wants to start coding.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, artifacts, posters, or applications (examples include websites, landing pages, dashboards, React components, HTML/CSS layouts, or when styling/beautifying any web UI). Generates creative, polished code and UI design that avoids generic AI aesthetics.