79 个 · 按相关度排序
AI Agent Workflow & Skill Architecture Guide — turn expert work into measurable, reusable agent systems. Covers workflow discovery, skill decomposition, harness design, evaluation, human escalation, observability, cost control, and multi-agent orchestration. Includes a measured marketing-delivery case: 15 people × 3–4 weeks reduced to one strategist + AI in 5 days. By Gingiris.
中文简介正在准备中。当前显示原文说明。
Sample skill that proves a DeepSeek Harness bundle imported into Codewhale. Use only when the user is trying the dsh-sample plugin.
中文简介正在准备中。当前显示原文说明。
Analyze one recorded agent session the way the Whalesong instrument does — signal shape first, computed findings second, exact evidence rows last. Use when the user asks what a session did, why it stalled, looped, errored, or burned tokens.
中文简介正在准备中。当前显示原文说明。
Review and rewrite the words in a user interface (buttons, labels, errors, empty states, confirmations, onboarding, notifications) for clarity, tone and consistency. Use when the user asks for UX writing feedback, microcopy, error messages, button text, or a content pass on a screen or flow.
中文简介正在准备中。当前显示原文说明。
Quantify periodicity in a recorded agent session — dominant onset period in seconds, spectral entropy, autocorrelation lag. Use to prove a poll loop, measure how metronomic retries are, compare cadence between runs, or window into one suspicious stretch.
中文简介正在准备中。当前显示原文说明。
Triage Linear issues and claim work with the Linear MCP tools - find the triage queue or your assigned work, check for duplicates and existing owners, set team/priority/labels, and claim an issue (assign to me, move to In Progress) only after confirming nobody else owns it. Use for "triage the queue", "what should I work on", "claim LIN-123".
中文简介正在准备中。当前显示原文说明。
Write an evidence-based status update from Linear data - summarize what shipped, what is in progress, what is blocked or stale for a project, cycle, or a person's issues, and optionally post it as a Linear project status update after the user approves the text. Use for "status for project X", "weekly update", "what moved this cycle".
中文简介正在准备中。当前显示原文说明。
Use Cloudflare KV, R2 and D1 from a Worker. Binding config, Worker code, cf commands, local vs remote data, D1 migrations. Use when adding or changing storage bindings or schemas.
中文简介正在准备中。当前显示原文说明。
Write a clear git commit message from the staged changes. Use when the user asks for a commit message, or to summarize what they are about to commit.
中文简介正在准备中。当前显示原文说明。
Triage and draft GitHub issues with the GitHub MCP tools - search for duplicates, classify and prioritize, and draft a well-formed new issue for the user to approve. Read-only until the user approves a specific create or label action. Use for "triage these issues", "is this a duplicate", or "file an issue for this bug".
中文简介正在准备中。当前显示原文说明。
Render a recorded agent session to audio with Whalesong's deterministic synthesizer and interpret it — hear a loop, a stall, an error storm, or the difference between two runs. Use when the user wants to feel a session's shape, compare runs by ear, or present a run without scrolling a waterfall.
中文简介正在准备中。当前显示原文说明。
Explain an unfamiliar repository, find where to make a change, and maintain source-cited documentation. Use for repo onboarding, wiki questions, documentation impact, or updating and sharing a WhaleWiki.
中文简介正在准备中。当前显示原文说明。
Attach review scores to recorded Whalesong traces and build comparison pairs — mark a session good/bad/regressed, then diff two runs deterministically. Use when evaluating agent changes, reviewing a run after the fact, or building a small eval set.
中文简介正在准备中。当前显示原文说明。
Sweep recent sessions across every agent CLI recorded in Whalesong and surface the ones worth a look — errors, runaway loops, token blowouts, stalled gaps. Use for "what have my agents been doing", morning reviews, or finding the session that went wrong.
中文简介正在准备中。当前显示原文说明。
Review a GitHub pull request with the GitHub MCP tools - read the description, diff, checks and existing discussion, then report findings in chat. Posts nothing to GitHub unless the user asks for that exact post and approves the text. Use for "review this PR", PR links, or "what changed in
中文简介正在准备中。当前显示原文说明。
Build, configure and test a Cloudflare Worker locally with cf (cloudflare.config.ts, typed bindings, static assets, cf dev, vitest). Use when creating or changing Worker code or configuration. Does not deploy; use cloudflare-deploy for that.
中文简介正在准备中。当前显示原文说明。
Build a self-contained web page, dashboard, prototype, calculator or small interactive tool as ONE HTML file (plain HTML/CSS/JS, or React plus Tailwind bundled into a single file), check it, and preview it locally. Use when the user wants a page they can open, share as a file, or iterate on in a browser.
中文简介正在准备中。当前显示原文说明。
Configure a Next.js app for Vercel. next.config versus vercel.json, proxy (renamed from middleware in Next.js 16), images, redirects and headers, function duration and region, caching and static export pitfalls, and local verification. Use when editing next.config, vercel.json or routing for a Next.js project on Vercel.
中文简介正在准备中。当前显示原文说明。
Debug a failing Vercel deployment or a misbehaving production site. Read build logs and runtime logs with vercel inspect and vercel logs, reproduce builds locally, find missing env vars and Node or output-directory mistakes. Read-only investigation.
中文简介正在准备中。当前显示原文说明。
Manage Vercel environment variables safely. Environments, vercel env ls/add/update/rm/pull/run, sensitive variables, NEXT_PUBLIC exposure, when changes take effect. Use when adding, rotating, auditing or pulling Vercel env vars.
中文简介正在准备中。当前显示原文说明。
Deploy a project to Vercel safely. Preview versus production, project linking, vercel CLI and Git-based flows, promote, rollback. Never deploys without the user's explicit approval. Use for any request to deploy, ship, publish, promote or roll back on Vercel.
中文简介正在准备中。当前显示原文说明。
Fan six specialist reviewers (general correctness, silent failures, type design, test gaps, comment accuracy, simplification) out over the current diff, a branch, or a PR in parallel, then merge their findings into one deduplicated, ranked report. Use for a thorough review before merging. Not for editing code or approving on someone's behalf.
中文简介正在准备中。当前显示原文说明。
Set up a bounded iterate-until-done run with the loop plugin's /loop command. Use when the user wants the same task retried or refined across several replies until a check passes, such as "keep going until the tests pass" or "do this up to 5 times".
中文简介正在准备中。当前显示原文说明。
Critique a user interface (screenshot, running page, HTML/CSS, mockup or description) and return severity-ranked findings with evidence and a concrete fix for each. Use when the user asks for design feedback, a UI review, "what's wrong with this screen", or a pre-launch polish pass.
中文简介正在准备中。当前显示原文说明。