[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EC38NTNTNZFSMZ1FTVXZ9W":3,"skill-same:01M7EC38NTNTNZFSMZ1FTVXZ9W":54},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M7EC38NTNTNZFSMZ1FTVXZ9W","cherry",null,"How to install, run, use and drive Cherry, the local web workspace (127.0.0.1) where Claude Code plans, documents and builds the projects in ~\u002FProjects. Use when the user mentions Cherry or the `cherry` command; asks to start, stop or open it; wants a project calibrated, surveyed, chatted with, scanned (security, tests, code health, dependencies) or handed to a background agent; asks for a scan report or what an agent did; wants to write a custom scan agent (~\u002F.cherry\u002Fagents\u002F*.md); edits _project\u002F notes the Cherry UI shows; or hits a Cherry problem (white screen, port 4317\u002F5173 in use, 401, \"Claude Code not found\").","syedmehedihussain\u002Fcherry","syedmehedihussain","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F183754033?v=4","01M7EC34EA312J2XV37SD5C840","Syed Mehedi Hussain","01M7EC34EZR59ARY7C75C41EKJ",3,"MIT","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcherry\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: cherry\ndescription: How to install, run, use and drive Cherry, the local web workspace (127.0.0.1) where Claude Code plans, documents and builds the projects in ~\u002FProjects. Use when the user mentions Cherry or the `cherry` command; asks to start, stop or open it; wants a project calibrated, surveyed, chatted with, scanned (security, tests, code health, dependencies) or handed to a background agent; asks for a scan report or what an agent did; wants to write a custom scan agent (~\u002F.cherry\u002Fagents\u002F*.md); edits _project\u002F notes the Cherry UI shows; or hits a Cherry problem (white screen, port 4317\u002F5173 in use, 401, \"Claude Code not found\").\n---\n\n# Cherry\n\nCherry is a local workspace where Claude Code works on the user's projects the way an engineer\nwould: survey first, decide, document, then build. A daemon (Fastify, `127.0.0.1:4317`) drives\nthe local `claude` CLI; a React UI talks to it. Nothing listens beyond loopback.\n\nTwo ways to help with it:\n\n- **Tell the user what to do in the UI**: the sections below say where everything is.\n- **Drive the running app yourself** through its API with `scripts\u002Fapi.sh` (next to this file).\n  Use this to start scans and agents, read reports, and check state. Never use it to answer\n  approvals or decisions; those are the user's (see Safety).\n\n## Install and run\n\nNeeds Node 22, pnpm, git, and Claude Code signed in (`claude auth status`). `gh` is optional\n(GitHub box, Push, new repos). The repo pins Node and pnpm in `mise.toml`.\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry && cd cherry\nmise install && pnpm install\n.\u002Fscripts\u002Finstall-cli.sh   # puts `cherry` in ~\u002F.local\u002Fbin\ncherry                    # start, print a login link, open the browser\n```\n\n| Command | Does |\n| --- | --- |\n| `cherry` \u002F `cherry up` | start the daemon (builds the UI if stale) and open the login link; `--no-open` only prints it |\n| `cherry start` \u002F `stop` \u002F `restart` | always on: a systemd user service that starts at login \u002F off \u002F restart after an update (ADR-0063) |\n| `cherry down` \u002F `status` | stop it until the next `cherry` or login \u002F say if it runs, how many projects, Claude status |\n| `cherry .` \u002F `cherry open \u003Cfolder>` | add a folder from anywhere to Home and open it (ADR-0031) |\n| `cherry open \u003Cproject>` | open that project's workspace |\n| `cherry doctor` | check Node, git, gh and claude |\n| `cherry logout` | end every browser (and api.sh) session |\n\nWorking on Cherry's own code: `pnpm dev` runs the daemon on 4317 and Vite on\n`http:\u002F\u002F127.0.0.1:5173` with live reload. It and `cherry` cannot run at the same time (same port,\nsame `~\u002F.cherry`); `pnpm dev` pauses the always-on service while it runs and starts it again after. The repo's `CLAUDE.md` has the dev rules.\n\n## Where things live\n\n| Path | What |\n| --- | --- |\n| `~\u002FProjects\u002F\u003Cname>\u002F` | one project per folder (the folder name is the project id); `projectsDir` in config changes the root |\n| `\u003Cproject>\u002F_project\u002F` | personal notes, git-excluded: `STATUS.md` (where we left off, next steps), `project.json` (phase, stack), `tasks.json`, `notes.md`, `survey.json`, `decisions.md` |\n| `\u003Cproject>\u002FCLAUDE.md`, `docs\u002F` | committed docs; the Docs tab lists them (ADRs in `docs\u002Fadr\u002F`) |\n| `\u003Cproject>\u002Fcherry\u002F` | git-ignored: `uploads\u002F` (chat attachments), `reports\u002F\u003Cagent>\u002F\u003Cdate>.md` (scan reports) |\n| `~\u002F.cherry\u002Fconfig.json` | the project list (`projects`), the new projects folder, default model, agent limits |\n| `~\u002F.cherry\u002Fagents\u002F*.md` | custom scan agents |\n| `~\u002F.cherry\u002Fmagnet\u002F` | Magnet's profile: `MAGNET.md`, `me.md`, `work.md`, generated `projects.md` |\n| `~\u002F.cherry\u002Frun\u002Fdaemon.json` | port, pid and CLI secret of the running daemon (0600) |\n| `\u003CprojectsDir>\u002F.cherry-worktrees\u002F` | git worktrees of background agents |\n\n## The UI\n\n**Home** lists projects by last worked, with state (Ready, cctop notes only, Not calibrated),\nphase and git. The prompt box starts a **new project survey** (seven steps, each a decision card,\nthen it writes CLAUDE.md, PRD, ADRs, STATUS and the first commit) or asks **Magnet**.\n\n**Calibration** (on a project that is not Ready): Claude reads the project read-only, asks a\nfew questions, then proposes files. Nothing is written until the user ticks the files; existing\nfiles only get additions.\n\n**Workspace** (`\u002Fp\u002F\u003Cproject>`), three columns:\n\n- Left: file tree with git letters (M, A, U, D) and coloured type tags; \"Claude touched\" filter.\n- Centre tabs: **Chat**, **Docs** (reader for CLAUDE.md, docs\u002F and ADRs), **Notes & Tasks**, the\n  open **file** (read-only, diff vs HEAD), and an open **report** (`\u002Fp\u002F\u003Cid>\u002Freports\u002F\u003Cagent>`).\n  The **Status** notch shows STATUS.md and can ask Claude to update it.\n- Right views: **Overview**, **Live**, **GitHub** (branch, pull, push, PRs, and Commits: stage and commit as you, or the **Auto commit** switch, where Claude commits its work through `commit_changes`, ADR-0030). Centre tabs: Chat, Docs, Notes & Tasks, Agents, Decisions.\n\n**Chat composer**: attach files (paperclip, drop or paste, max 8 x 10 MB), **Plan mode** (Claude\nplans, no edits), model (Sonnet, Opus, Haiku), **Effort** (Default, Low to Max; hidden for\nHaiku). Every edit and command shows an **approval card**: Allow, Allow for this session, Deny.\nClaude may ask a **decision card** with options and a recommendation.\n\n**Agents view**:\n\n- **Scans**: one-click read-only agents that write a report. Built in: Security review (Opus),\n  Test runner, Code health, Dependency audit (Sonnet). They never ask anything: read tools run,\n  edits and the web are off, and only the agent's own commands run (test\u002Flint\u002Ftypecheck for the\n  Test runner, audit\u002Foutdated for Dependency audit, read-only git for all). The report opens in\n  the centre with the history of earlier runs. Dependency audit contacts the package registry.\n- **Tasks**: a background agent gets a free-text task and works in its own worktree on\n  `agent\u002F\u003Cslug>`. It always runs in Auto: edits and commands need no approval. Push, publish,\n  deploy and `rm -r` are skipped and listed as **Needs you**; **Continue in chat** hands them to\n  the chat (ADR-0028). When done: Review diff, then Merge (into the current branch; on a conflict\n  **Merge with agent** lets the agent fix it) or Discard. Needs a git repo with one commit.\n\n**Magnet** is the cross-project assistant: read-only across all projects; it can propose actions\n(start an agent, calibrate, new project) that run only after the user approves.\n\n## Driving Cherry with scripts\u002Fapi.sh\n\n`scripts\u002Fapi.sh METHOD PATH [JSON]` prints the JSON response; exit 1 on an HTTP error, 2 when\nCherry is not running. It signs in with the CLI secret and keeps a cookie in\n`~\u002F.cherry\u002Frun\u002Fskill-cookies.txt`. Run it by its full path, e.g.\n`.claude\u002Fskills\u002Fcherry\u002Fscripts\u002Fapi.sh` inside the repo or `~\u002F.claude\u002Fskills\u002Fcherry\u002Fscripts\u002Fapi.sh`\nwhen installed for the user.\n\n```bash\napi.sh GET  \u002Fapi\u002Fhealth                          # Claude found and signed in?\napi.sh GET  \u002Fapi\u002Fprojects                        # project cards (id, state, phase, next step)\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>                   # card, STATUS.md, docs, git\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fscans             # scan agents, running run, last report\napi.sh POST \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fscans\u002Fsecurity\u002Frun\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Freports\u002Fsecurity  # history, newest first\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Freports\u002Fsecurity\u002F\u003CreportId>   # { meta, markdown }\napi.sh POST \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fagents '{\"task\":\"Write tests for the streak service\",\"model\":\"sonnet\"}'\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fagents            # status, summary, waiting approval\napi.sh GET  \u002Fapi\u002Fagents\u002F\u003CagentId>\u002Fdiff\napi.sh GET  \u002Fapi\u002Fapprovals?status=pending\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fchat              # transcript of the current chat\n```\n\nA scan takes one to several minutes. Poll `GET ...\u002Fscans` until that agent's `run` is `null`,\nthen read `last` and fetch the report. Reports are also plain files under\n`\u003Cproject>\u002Fcherry\u002Freports\u002F`, so reading them from disk works too. The full API is in the repo's\n`docs\u002Fapi.md`; request bodies are the zod schemas in `packages\u002Fshared\u002Fsrc\u002Fschemas\u002F`.\n\n## Safety\n\nThe approval cards are the user's control over what Claude does to their files. When driving\nthe API:\n\n- **Never** `POST \u002Fapi\u002Fapprovals\u002F:id`, `POST \u002Fapi\u002Fdecisions\u002F:id`, `...\u002Fcalibrate\u002Fwrite`,\n  `\u002Fapi\u002Fagents\u002F:id\u002Faccept`, `...\u002Fgit\u002Fpush` or `\u002Fapi\u002Fmagnet\u002Factions\u002F:id\u002Fapprove` unless the user\n  explicitly asked for that exact action in this conversation. Tell them what is waiting and\n  where to click instead.\n- Starting a scan, reading reports, and reading state are fine when they fit the request.\n  Starting a background agent or a chat turn uses the user's Claude plan: say so first.\n- Do not `PATCH \u002Fapi\u002Fconfig` or edit `~\u002F.cherry\u002F` without asking.\n- Never print `daemon.json` or the cookie file; they are credentials.\n\n## Writing a custom scan agent\n\nCreate `~\u002F.cherry\u002Fagents\u002F\u003Cid>.md` (the file name is the id; an id equal to a built-in replaces\nit). It shows up under Scans with a \"custom\" tag; a broken file shows its error there.\n\n```markdown\n---\nname: Accessibility check\ndescription: Labels, alt text, keyboard traps and contrast\nicon: shield            # shield | flask | heart | package | bot\nmodel: sonnet           # sonnet | opus | haiku\ncommands: pnpm lint, pnpm test   # optional; exact command or it plus arguments\n---\nWhat to look for, in what order, and how to rate severity (critical, high, medium, low).\nCherry adds the read-only rules and the report format (Summary, Findings, What to do next).\n```\n\n## Reports are in ASD-STE100\n\nEvery report that an agent or subagent writes is in ASD-STE100 Simplified Technical English.\nThis applies to Cherry scan reports, background agent summaries, custom scan agents, and any\nsubagent you start (Agent tool) while you work with Cherry. Tell each subagent this in its prompt.\n\n- Short sentences: max 20 words in a procedure, max 25 in a description.\n- One instruction per sentence. Start instructions with the verb (\"Run the tests.\").\n- Active voice. Use approved words, with one meaning for each word.\n- Keep code, paths, commands, product names and error text exactly as they are.\n\nCherry adds this rule to the built-in scan and agent prompts. A custom scan agent gets it from\nthe report format, so its body does not need to repeat it.\n\n## Troubleshooting\n\n| Symptom | Fix |\n| --- | --- |\n| `Port 5173 is already in use` \u002F `4317` | another `pnpm dev` or `cherry` runs: `cherry down`, or find it with `ss -ltnp \\| grep -E '4317\\|5173'` and stop that process |\n| White screen in dev, console says a module \"does not provide an export\" | Vite cached a half-written file (often after a bulk `prettier --write`); `touch` that file and reload |\n| `401 unauthorized` in the browser | run `cherry` and open the link it prints (codes are single-use, 10 minutes) |\n| \"Claude Code not found\" \u002F not signed in | install Claude Code, run `claude auth login`, then \"Check again\" or `cherry doctor` |\n| Agent: \"Agents need a git repository\" \u002F \"Make a first commit\" | `git init` and commit once in the project |\n| Scan ends \"Did not finish\" | open the report for the error; usually Claude sign-in or a stop; run it again |\n| Project missing on Home | Cherry lists only opened or created projects: **Open folder…** on Home or `cherry .` in the folder (ADR-0032) |\n","# Cherry\n\nCherry is a local workspace where Claude Code works on the user's projects the way an engineer\nwould: survey first, decide, document, then build. A daemon (Fastify, `127.0.0.1:4317`) drives\nthe local `claude` CLI; a React UI talks to it. Nothing listens beyond loopback.\n\nTwo ways to help with it:\n\n- **Tell the user what to do in the UI**: the sections below say where everything is.\n- **Drive the running app yourself** through its API with `scripts\u002Fapi.sh` (next to this file).\n  Use this to start scans and agents, read reports, and check state. Never use it to answer\n  approvals or decisions; those are the user's (see Safety).\n\n## Install and run\n\nNeeds Node 22, pnpm, git, and Claude Code signed in (`claude auth status`). `gh` is optional\n(GitHub box, Push, new repos). The repo pins Node and pnpm in `mise.toml`.\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry && cd cherry\nmise install && pnpm install\n.\u002Fscripts\u002Finstall-cli.sh   # puts `cherry` in ~\u002F.local\u002Fbin\ncherry                    # start, print a login link, open the browser\n```\n\n| Command | Does |\n| --- | --- |\n| `cherry` \u002F `cherry up` | start the daemon (builds the UI if stale) and open the login link; `--no-open` only prints it |\n| `cherry start` \u002F `stop` \u002F `restart` | always on: a systemd user service that starts at login \u002F off \u002F restart after an update (ADR-0063) |\n| `cherry down` \u002F `status` | stop it until the next `cherry` or login \u002F say if it runs, how many projects, Claude status |\n| `cherry .` \u002F `cherry open \u003Cfolder>` | add a folder from anywhere to Home and open it (ADR-0031) |\n| `cherry open \u003Cproject>` | open that project's workspace |\n| `cherry doctor` | check Node, git, gh and claude |\n| `cherry logout` | end every browser (and api.sh) session |\n\nWorking on Cherry's own code: `pnpm dev` runs the daemon on 4317 and Vite on\n`http:\u002F\u002F127.0.0.1:5173` with live reload. It and `cherry` cannot run at the same time (same port,\nsame `~\u002F.cherry`); `pnpm dev` pauses the always-on service while it runs and starts it again after. The repo's `CLAUDE.md` has the dev rules.\n\n## Where things live\n\n| Path | What |\n| --- | --- |\n| `~\u002FProjects\u002F\u003Cname>\u002F` | one project per folder (the folder name is the project id); `projectsDir` in config changes the root |\n| `\u003Cproject>\u002F_project\u002F` | personal notes, git-excluded: `STATUS.md` (where we left off, next steps), `project.json` (phase, stack), `tasks.json`, `notes.md`, `survey.json`, `decisions.md` |\n| `\u003Cproject>\u002FCLAUDE.md`, `docs\u002F` | committed docs; the Docs tab lists them (ADRs in `docs\u002Fadr\u002F`) |\n| `\u003Cproject>\u002Fcherry\u002F` | git-ignored: `uploads\u002F` (chat attachments), `reports\u002F\u003Cagent>\u002F\u003Cdate>.md` (scan reports) |\n| `~\u002F.cherry\u002Fconfig.json` | the project list (`projects`), the new projects folder, default model, agent limits |\n| `~\u002F.cherry\u002Fagents\u002F*.md` | custom scan agents |\n| `~\u002F.cherry\u002Fmagnet\u002F` | Magnet's profile: `MAGNET.md`, `me.md`, `work.md`, generated `projects.md` |\n| `~\u002F.cherry\u002Frun\u002Fdaemon.json` | port, pid and CLI secret of the running daemon (0600) |\n| `\u003CprojectsDir>\u002F.cherry-worktrees\u002F` | git worktrees of background agents |\n\n## The UI\n\n**Home** lists projects by last worked, with state (Ready, cctop notes only, Not calibrated),\nphase and git. The prompt box starts a **new project survey** (seven steps, each a decision card,\nthen it writes CLAUDE.md, PRD, ADRs, STATUS and the first commit) or asks **Magnet**.\n\n**Calibration** (on a project that is not Ready): Claude reads the project read-only, asks a\nfew questions, then proposes files. Nothing is written until the user ticks the files; existing\nfiles only get additions.\n\n**Workspace** (`\u002Fp\u002F\u003Cproject>`), three columns:\n\n- Left: file tree with git letters (M, A, U, D) and coloured type tags; \"Claude touched\" filter.\n- Centre tabs: **Chat**, **Docs** (reader for CLAUDE.md, docs\u002F and ADRs), **Notes & Tasks**, the\n  open **file** (read-only, diff vs HEAD), and an open **report** (`\u002Fp\u002F\u003Cid>\u002Freports\u002F\u003Cagent>`).\n  The **Status** notch shows STATUS.md and can ask Claude to update it.\n- Right views: **Overview**, **Live**, **GitHub** (branch, pull, push, PRs, and Commits: stage and commit as you, or the **Auto commit** switch, where Claude commits its work through `commit_changes`, ADR-0030). Centre tabs: Chat, Docs, Notes & Tasks, Agents, Decisions.\n\n**Chat composer**: attach files (paperclip, drop or paste, max 8 x 10 MB), **Plan mode** (Claude\nplans, no edits), model (Sonnet, Opus, Haiku), **Effort** (Default, Low to Max; hidden for\nHaiku). Every edit and command shows an **approval card**: Allow, Allow for this session, Deny.\nClaude may ask a **decision card** with options and a recommendation.\n\n**Agents view**:\n\n- **Scans**: one-click read-only agents that write a report. Built in: Security review (Opus),\n  Test runner, Code health, Dependency audit (Sonnet). They never ask anything: read tools run,\n  edits and the web are off, and only the agent's own commands run (test\u002Flint\u002Ftypecheck for the\n  Test runner, audit\u002Foutdated for Dependency audit, read-only git for all). The report opens in\n  the centre with the history of earlier runs. Dependency audit contacts the package registry.\n- **Tasks**: a background agent gets a free-text task and works in its own worktree on\n  `agent\u002F\u003Cslug>`. It always runs in Auto: edits and commands need no approval. Push, publish,\n  deploy and `rm -r` are skipped and listed as **Needs you**; **Continue in chat** hands them to\n  the chat (ADR-0028). When done: Review diff, then Merge (into the current branch; on a conflict\n  **Merge with agent** lets the agent fix it) or Discard. Needs a git repo with one commit.\n\n**Magnet** is the cross-project assistant: read-only across all projects; it can propose actions\n(start an agent, calibrate, new project) that run only after the user approves.\n\n## Driving Cherry with scripts\u002Fapi.sh\n\n`scripts\u002Fapi.sh METHOD PATH [JSON]` prints the JSON response; exit 1 on an HTTP error, 2 when\nCherry is not running. It signs in with the CLI secret and keeps a cookie in\n`~\u002F.cherry\u002Frun\u002Fskill-cookies.txt`. Run it by its full path, e.g.\n`.claude\u002Fskills\u002Fcherry\u002Fscripts\u002Fapi.sh` inside the repo or `~\u002F.claude\u002Fskills\u002Fcherry\u002Fscripts\u002Fapi.sh`\nwhen installed for the user.\n\n```bash\napi.sh GET  \u002Fapi\u002Fhealth                          # Claude found and signed in?\napi.sh GET  \u002Fapi\u002Fprojects                        # project cards (id, state, phase, next step)\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>                   # card, STATUS.md, docs, git\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fscans             # scan agents, running run, last report\napi.sh POST \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fscans\u002Fsecurity\u002Frun\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Freports\u002Fsecurity  # history, newest first\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Freports\u002Fsecurity\u002F\u003CreportId>   # { meta, markdown }\napi.sh POST \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fagents '{\"task\":\"Write tests for the streak service\",\"model\":\"sonnet\"}'\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fagents            # status, summary, waiting approval\napi.sh GET  \u002Fapi\u002Fagents\u002F\u003CagentId>\u002Fdiff\napi.sh GET  \u002Fapi\u002Fapprovals?status=pending\napi.sh GET  \u002Fapi\u002Fprojects\u002F\u003Cid>\u002Fchat              # transcript of the current chat\n```\n\nA scan takes one to several minutes. Poll `GET ...\u002Fscans` until that agent's `run` is `null`,\nthen read `last` and fetch the report. Reports are also plain files under\n`\u003Cproject>\u002Fcherry\u002Freports\u002F`, so reading them from disk works too. The full API is in the repo's\n`docs\u002Fapi.md`; request bodies are the zod schemas in `packages\u002Fshared\u002Fsrc\u002Fschemas\u002F`.\n\n## Safety\n\nThe approval cards are the user's control over what Claude does to their files. When driving\nthe API:\n\n- **Never** `POST \u002Fapi\u002Fapprovals\u002F:id`, `POST \u002Fapi\u002Fdecisions\u002F:id`, `...\u002Fcalibrate\u002Fwrite`,\n  `\u002Fapi\u002Fagents\u002F:id\u002Faccept`, `...\u002Fgit\u002Fpush` or `\u002Fapi\u002Fmagnet\u002Factions\u002F:id\u002Fapprove` unless the user\n  explicitly asked for that exact action in this conversation. Tell them what is waiting and\n  where to click instead.\n- Starting a scan, reading reports, and reading state are fine when they fit the request.\n  Starting a background agent or a chat turn uses the user's Claude plan: say so first.\n- Do not `PATCH \u002Fapi\u002Fconfig` or edit `~\u002F.cherry\u002F` without asking.\n- Never print `daemon.json` or the cookie file; they are credentials.\n\n## Writing a custom scan agent\n\nCreate `~\u002F.cherry\u002Fagents\u002F\u003Cid>.md` (the file name is the id; an id equal to a built-in replaces\nit). It shows up under Scans with a \"custom\" tag; a broken file shows its error there.\n\n```markdown\n---\nname: Accessibility check\ndescription: Labels, alt text, keyboard traps and contrast\nicon: shield            # shield | flask | heart | package | bot\nmodel: sonnet           # sonnet | opus | haiku\ncommands: pnpm lint, pnpm test   # optional; exact command or it plus arguments\n---\nWhat to look for, in what order, and how to rate severity (critical, high, medium, low).\nCherry adds the read-only rules and the report format (Summary, Findings, What to do next).\n```\n\n## Reports are in ASD-STE100\n\nEvery report that an agent or subagent writes is in ASD-STE100 Simplified Technical English.\nThis applies to Cherry scan reports, background agent summaries, custom scan agents, and any\nsubagent you start (Agent tool) while you work with Cherry. Tell each subagent this in its prompt.\n\n- Short sentences: max 20 words in a procedure, max 25 in a description.\n- One instruction per sentence. Start instructions with the verb (\"Run the tests.\").\n- Active voice. Use approved words, with one meaning for each word.\n- Keep code, paths, commands, product names and error text exactly as they are.\n\nCherry adds this rule to the built-in scan and agent prompts. A custom scan agent gets it from\nthe report format, so its body does not need to repeat it.\n\n## Troubleshooting\n\n| Symptom | Fix |\n| --- | --- |\n| `Port 5173 is already in use` \u002F `4317` | another `pnpm dev` or `cherry` runs: `cherry down`, or find it with `ss -ltnp \\| grep -E '4317\\|5173'` and stop that process |\n| White screen in dev, console says a module \"does not provide an export\" | Vite cached a half-written file (often after a bulk `prettier --write`); `touch` that file and reload |\n| `401 unauthorized` in the browser | run `cherry` and open the link it prints (codes are single-use, 10 minutes) |\n| \"Claude Code not found\" \u002F not signed in | install Claude Code, run `claude auth login`, then \"Check again\" or `cherry doctor` |\n| Agent: \"Agents need a git repository\" \u002F \"Make a first commit\" | `git init` and commit once in the project |\n| Scan ends \"Did not finish\" | open the report for the error; usually Claude sign-in or a stop; run it again |\n| Project missing on Home | Cherry lists only opened or created projects: **Open folder…** on Home or `cherry .` in the folder (ADR-0032) |\n",{"name":5,"description":7},"main","adb38ff1e06b1bb57c84131717dd3d9c01292a8b",".claude\u002Fskills\u002Fcherry",0,"https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry","https:\u002F\u002Fsyedmehedihussain.github.io\u002Fcherry\u002F","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain","https:\u002F\u002Fsyedmehedihussain.codes\u002F","What should I say?","2026-11-10 02:42:31",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add syedmehedihussain\u002Fcherry --skill cherry",{"tool":40,"command":41},"bunx","bunx skills add syedmehedihussain\u002Fcherry --skill cherry",{"tool":43,"command":44},"pnpm","pnpm dlx skills add syedmehedihussain\u002Fcherry --skill cherry","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcherry\u002FSKILL.md",true,[48],{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"scripts\u002Fapi.sh","SCRIPT","脚本",2107,"https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcherry\u002Fscripts\u002Fapi.sh",[55,63,71,79,87,95],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":59,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7ECBMEN6N02N2NGHMYTJTY1","verification-loop","Run a six-phase verification of a Claude Code session's work — build, type check, lint, tests with coverage, security grep, and diff review — then produce a PASS\u002FFAIL verification report. Use when verifying work after completing a feature or refactor, before creating a PR, or when quality gates must pass.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fverification-loop\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fverification-loop\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":67,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7ECBFBZ3GJEWJJHKRPCYX8J","e2e-testing","Playwright E2E testing patterns, Page Object Model, configuration, CI\u002FCD integration, artifact management, and flaky test strategies. Use when writing Playwright tests, structuring page objects, or fixing flaky E2E runs in CI.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fe2e-testing\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fe2e-testing\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":75,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":19},"01M7ECB8WD04GN1SF0J7XM0252","browser-qa","Run automated post-deploy UI verification with a browser automation MCP (claude-in-chrome, Playwright, or Puppeteer): console-error and Core Web Vitals smoke checks, form and auth-flow interaction tests, screenshot visual regression across three breakpoints, and axe-core accessibility audits ending in a SHIP \u002F DO-NOT-SHIP verdict. Use when testing a deployed feature on staging or preview, before shipping frontend changes, reviewing a frontend PR, or checking responsive layout and accessibility.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fbrowser-qa\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fbrowser-qa\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":83,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":19},"01M7ECB2QHYKHYT9EJV089FVT4","ai-regression-testing","Regression testing strategies for AI-assisted development. Sandbox-mode API testing without database dependencies, automated bug-check workflows, and patterns to catch AI blind spots where the same model writes and reviews code. Use when adding regression coverage to AI-assisted code, or when the same model both wrote and reviewed a change.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fai-regression-testing\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Ftesting\u002Fskills\u002Fai-regression-testing\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":91,"sourceHtmlUrl":92,"categories":93,"keywords":94,"publishedAt":19},"01M7ECAXS1W69TNH9MV1J1WHM2","story-lifecycle","Manage a file-based epic → story → sprint development loop without requiring an external project management tool. Creates and tracks epics, breaks them into user stories, assigns them to sprints, and drives implementation story by story. Use when starting a new feature initiative or when the team has no Jira\u002FLinear\u002FTaiga but needs structured delivery tracking.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Frequirements\u002Fskills\u002Fstory-lifecycle\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Frequirements\u002Fskills\u002Fstory-lifecycle\u002FSKILL.md",[],[],{"skillNo":96,"name":97,"nameZh":6,"description":98,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":99,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":19},"01M7ECAQVHTTS6M1T62XYAP9NP","intent-driven-development","Turn ambiguous or high-impact product and engineering changes into scoped, verifiable acceptance criteria before or alongside implementation. Use when a user asks to clarify a feature, define acceptance criteria, de-risk a security\u002Fdata\u002Fmigration\u002Fintegration change, prepare implementation requirements for another agent, or make a complex request testable. Do not trigger for trivial edits, straightforward fixes, active debugging, code review, or implementation requests whose acceptance conditions are already clear unless the user explicitly invokes this skill.","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Frequirements\u002Fskills\u002Fintent-driven-development\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fsyedmehedihussain\u002Fcherry\u002Fblob\u002Fmain\u002Fpackages\u002Fdaemon\u002Fstages\u002Frequirements\u002Fskills\u002Fintent-driven-development\u002FSKILL.md",[],[]]