[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7CJY49MS2XBY45MWJBMM3FW":3,"skill-same:01M7CJY49MS2XBY45MWJBMM3FW":65},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7CJY49MS2XBY45MWJBMM3FW","master-of",null,"Cross-domain entry point for the master-of skill-gate system: use this when a request needs dormant skills from SEVERAL domains at once (e.g. scrape a competitor AND critique its design, review a UI AND its copy) and you'd otherwise have to invoke \u002Fdesign, \u002Fresearch, \u002Fdev, \u002Fstock, \u002Fplanning, \u002Fpipelines one at a time. Also the entry point for any category that has no gate of its own (one added later via check-skills). Matches across every category and activates the 1-4 skills that fit, from however many domains. For a request that clearly sits in one of the six built-in domains, invoke that single gate instead; for management\u002Fstatus (scanning, classifying, reporting), invoke 'check-skills'.","devrudals\u002Fmaster-of","devrudals","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F229339820?v=4","01M7CJY2T414404AVZ7E87C9J9","01M7CJY2TRFSPRG6QNY2VE3130",3,"MIT","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: master-of\ndescription: \"Cross-domain entry point for the master-of skill-gate system: use this when a request needs dormant skills from SEVERAL domains at once (e.g. scrape a competitor AND critique its design, review a UI AND its copy) and you'd otherwise have to invoke \u002Fdesign, \u002Fresearch, \u002Fdev, \u002Fstock, \u002Fplanning, \u002Fpipelines one at a time. Also the entry point for any category that has no gate of its own (one added later via check-skills). Matches across every category and activates the 1-4 skills that fit, from however many domains. For a request that clearly sits in one of the six built-in domains, invoke that single gate instead; for management\u002Fstatus (scanning, classifying, reporting), invoke 'check-skills'.\"\nallowed-tools: Read\n---\n\n# master-of — overview\n\nThis plugin bundles:\n- **Domain gates**: `design` \u002F `dev` \u002F `research` \u002F\n  `stock` \u002F `planning` — on-demand activation for a dormant\n  skill in that domain. Each is a thin pointer to the Activation Protocol\n  below.\n- **`check-skills`** — the system's own manager. It scans for\n  new\u002Fremoved skills and plugins (automatically via the `SessionStart` hook,\n  or on demand), classifies new finds into the right gate (or a new one),\n  and reports the current full breakdown read-only. See\n  `.\u002Fskills\u002Fcheck-skills\u002FSKILL.md` for everything about how that works —\n  this file doesn't duplicate it.\n\n## Activation protocol (shared by every domain gate)\n\nEvery domain gate (`design`\u002F`dev`\u002F`research`\u002F`stock`\u002F`planning`) delegates its\nactual behavior here instead of repeating it — each gate's own SKILL.md is\njust a pointer to this section plus its category key and any\ncategory-specific notes.\n\n**When to open a gate at all.** Two ways in, both fine: the user invokes it\n(`\u002Fdesign …`), or you open it yourself because the user asked for actual\n*work* that clearly belongs to that domain (\"이 화면 다듬어줘\", \"경쟁사\n스크래핑해줘\"). Don't open one for a question, an explanation, or chat that\nmerely mentions the topic — opening costs tokens (100-2,800 per gate), so it\nhas to be a task you're about to do. When in doubt, ask in one line rather\nthan opening. Once a gate is open:\n\n1. Read `~\u002F.claude\u002Fmasterof\u002Fgates\u002F\u003Ccategory>.txt` — the pre-rendered index for\n   that one gate (`name | 설명 | Read할 경로`, one line per entry).\n   **Never `Read` `~\u002F.claude\u002Fmasterof\u002Fregistry.json` to activate a gate.**\n   `Read` pulls the whole file, and registry.json is ~60KB \u002F ~16k tokens —\n   more than 2.5x the entire per-session saving this system buys, spent on a\n   single activation. The gate files carry the same name\u002Fdescription\u002Fpath the\n   protocol needs, at 100-2,500 tokens each. registry.json is the source of\n   truth for *classification* (`check-skills` writes it); the `gates\u002F`\n   files are the read path for *activation*, regenerated from it by\n   `scripts\u002Frender-report.ts` on every change.\n2. **Also glance at scale — the related pipelines are already in the file.**\n   Judge whether the request is a narrow, specific task (→ stay with the\n   domain entries) or something project-\u002Fworkflow-scale (\"전체 다 해줘\",\n   \"처음부터 끝까지\", a whole new design direction, a whole phase cycle) that a\n   complete pipeline would handle better. Each gate file ends with a\n   `## 관련 파이프라인` section listing exactly the pipelines whose `domain`\n   is this category, so this costs no extra read. Don't treat `design` and\n   `pipelines` as separate lanes the user has to know to pick between; a\n   design-shaped request should surface a fitting pipeline (e.g. `paint`) as\n   naturally as it surfaces a supporting skill (e.g. `better-ui`), when the\n   scale calls for it. If both a narrow skill and a pipeline plausibly fit,\n   say so and let the user choose which scale they want, rather than guessing.\n3. **Try to match first — don't show anything yet.** Compare what the user\n   actually asked for (the message that triggered this skill, or the task at\n   hand) against each entry's `description`.\n   - One confident match → say so directly (\"이 작업엔 `\u003Cname>`을 쓸게요\") and\n     `Read` its `path`. No list, no round-trip.\n   - A small confident cluster (2-4 entries that clearly belong together for\n     this one task) → name them directly the same way and proceed.\n   - **Genuinely ambiguous** (several entries plausibly fit and none clearly\n     wins) → check `~\u002F.claude\u002Fmasterof\u002Fpreferences.json` for this category\n     before asking:\n     - `{\"mode\": \"always_ask\"}` → print the full list (grouped by `cluster`\n       for `planning`, flat for the others) and ask, offering\n       `전체`\u002F`all`, numbers, names, or (for `planning`) a cluster name.\n     - `{\"mode\": \"fixed_default\", \"default\": \"\u003Cname>\"}` → use that entry\n       without asking, but say which one and that it's the saved default (so\n       the user can override in the moment if it's wrong for this task).\n     - `{\"mode\": \"conditional\", \"rules\": [{\"when\": \"\u003Ccondition, in plain\n       language>\", \"use\": \"\u003Cname>\"}, ...], \"fallback\": \"always_ask\" |\n       \"\u003Cname>\"}` → check each rule's `when` against the task in order (plain\n       judgment, not string matching); use the first match's `use`; if none\n       match, fall back to `fallback` (either ask, or another fixed default).\n     - `{\"mode\": \"smart\"}` → judge from context yourself and ask only when\n       truly tied.\n   - If the user explicitly asks something like \"뭐 있어\" \u002F \"다 보여줘\" \u002F \"전체\n     목록\", always show the full list regardless of preference mode — that's\n     an explicit request to see everything, not an ambiguous match.\n4. `Read` only what got selected — never the whole category \"just in case.\"\n   Bulk-reading every entry in a category on every invocation is exactly the\n   always-on cost this system exists to avoid; a gate that does that is a bug,\n   not a feature.\n   **Carry the base directory with it.** A gated skill is activated by\n   `Read`ing its `SKILL.md` directly, which — unlike a real `Skill` invocation\n   — supplies no \"Base directory for this skill\" line. Many of these skills\n   ship companion files and refer to them relatively (`RECIPES.md`,\n   `STANDARDS.md`, `reference\u002F`, `scripts\u002F…`); with no anchor those references\n   are dead ends. So treat the directory containing the `SKILL.md` you just\n   read (the path in the gate index, minus the filename) as that skill's base\n   directory, and resolve every relative path in its body against it. State it\n   once when you activate (\"base: `…\u002Fskills-library\u002Fdesign\u002Fanimate\u002F`\") so it's\n   explicit rather than re-derived each time you follow a reference.\n   **Reading a skill is not the same as invoking it — three things the\n   `Skill` tool would have done for you that you now do yourself:**\n   - `$ARGUMENTS` is not substituted. Many skills (every GSD skill, for\n     one) say things like \"Phase number: $ARGUMENTS\". Treat whatever the\n     user gave after the skill's name — or the specific thing they asked for\n     — as that value, and if the skill needs an argument the user didn't\n     supply, ask for it the way the skill itself says to (most say what to\n     do when it's omitted).\n   - `allowed-tools` in the frontmatter doesn't pre-approve anything. The\n     skill will work; you may just get permission prompts it wouldn't\n     normally trigger. Don't read that as the skill being broken.\n   - `disable-model-invocation: true` in the frontmatter means the author\n     wants it run only when the user explicitly asks for it BY NAME — never\n     on your own judgment. Opening a gate yourself (the auto-open rule) does\n     not lift that: if the best match carries this flag and the user didn't\n     name it, mention it as an option and let them choose; don't activate it.\n   **Commands and agents are gated the same way, and the index tags them.**\n   Plugins ship more than skills, and every command and agent costs\n   always-on tokens exactly like a skill does, so they're classified into\n   the same gates. An index line starting `[커맨드]` or `[에이전트]` is\n   activated differently from a skill:\n   - `[커맨드]` → `Read` it. The body is the prompt a `\u002Fname args` invocation\n     would have run; `$ARGUMENTS` is the user's request (same rule as\n     above). Follow it as those instructions, base directory rule included.\n   - `[에이전트]` → `Read` it. The frontmatter's `description`\u002F`tools`\u002F\n     `model` and the body (its system prompt) are what the harness would\n     have used to register it as a `subagent_type`. Since it isn't\n     registered, spawn it yourself: `Agent` with `subagent_type:\n     \"general-purpose\"`, `prompt` = the body followed by the concrete task,\n     `model` = the frontmatter's if it names one you can pass, and the\n     `tools` line restated inside the prompt as a constraint (\"only use\n     …\"). Say once that you're running it this way.\n   - **A gated skill that says \"spawn `subagent_type: X`\"** where X isn't in\n     the available agent types is *not* broken: check the current gate's\n     index (then `_all.txt`) for `[에이전트] X` and run it as above. This is\n     the normal path for frameworks whose skills and agents were gated\n     together. It costs one `Read` of X's file per spawn on your side — a\n     framework that spawns many agents per workflow pays that each time,\n     which is why `check-skills` asks before gating a large agent set.\n   **Honor `[의존: …]` notes.** An index line can carry\n   `[의존: \u003Cplugin> \u003Ccomponent> — 현재 꺼짐, 없으면 동작 불가]`. That means the\n   skill's plugin is disabled, so the MCP server \u002F agents it needs don't\n   exist this session. If it says 없으면 동작 불가, don't activate it — tell\n   the user it can't work right now and what would fix it (`claude plugin\n   enable \u003Cplugin>`, effective next session), and offer the closest working\n   alternative instead. Without that suffix (a partial), activate but say\n   which parts won't run. The status is as of the last render; the session-\n   start hook is the fresh source and will already have said so if it's off.\n5. **A match is a candidate, not an obligation.** The index answers \"what\n   exists in this domain\", which is not the same question as \"does this task\n   actually need it\". If the best match is only topically adjacent — it shares\n   the subject but not the job (a \"how to build X\" guide when the task is\n   \"decide whether to build X\"; a reviewer skill when the task is to gather\n   facts) — say so and skip the Read instead of activating it for form's sake.\n   Reading a 2-3k token skill that doesn't change what you do costs more than\n   the gate ever saved.\n6. Briefly confirm what's active, then proceed with the actual task.\n\nScope: one-shot, in-context only. Nothing persists, nothing is written to\nsettings. Re-invoke the gate for the next unrelated task in that domain.\n\n### Multi-activation — several skills, and several domains, at once\n\nActivation is **not** limited to one skill. A real request often needs two or\nthree things working together (\"이 화면 리뷰하고 접근성까지 봐줘\", \"경쟁사\n스크래핑해서 그 디자인 분석해줘\"), and forcing the user to invoke gates one at a\ntime, serially, is the single most annoying way this system can behave.\n\n**Within one domain** — step 3 already covers it: a confident cluster of 2-4\nentries is named and read together, not one at a time.\n\n**Across domains** — same idea, one level up:\n\n- If you can name the domains from the request (usually you can — \"스크래핑\"\n  → `research`, \"디자인 분석\" → `design`), read just those gate files. Two\n  gate files is typically 1-2k tokens total, far cheaper than the combined\n  index.\n- Only when the domain set is genuinely unclear, read\n  `~\u002F.claude\u002Fmasterof\u002Fgates\u002F_all.txt` — every category in one file (~6k\n  tokens). It's the cross-domain fallback, not the default.\n- Then match within the combined set and activate what fits, naming each one\n  and which domain it came from (\"`firecrawl-scrape`(research) + `better-ui`\n  (design) 두 개 붙일게요\"). Read only those.\n- **Cap it at ~4 skills.** Beyond that the activations start crowding out the\n  actual task, and it's a sign the request should be split — or that a single\n  `pipelines` entry is the right answer instead of five supporting skills.\n- The `pipelines` single-select rule still holds across domains: at most ONE\n  pipeline, and a pipeline generally shouldn't be combined with supporting\n  skills from other domains — it already runs its own end-to-end workflow.\n\nInvoking the root `master-of` skill directly (`\u002Fmaster-of \u003C요청>`) is the\nexplicit cross-domain entry point: it runs exactly this step — match across\nevery category first, then activate the 1-4 skills that fit, from however\nmany domains they happen to live in.\n\n### Staying activated across a continued task\n\nDon't re-run the full matching step from scratch on every single message.\nWhen a new user request comes in while something is already active from this\ndomain:\n- **Same or clearly continuing task** → keep using what's already active, no\n  re-invocation needed. Do a quick silent check whether a different entry in\n  the category would now fit better (the task may have shifted); if so,\n  mention it as a one-line suggestion rather than silently switching.\n- **A new, different task in the same domain** → re-run the matching step\n  fresh (steps 1-4 above); it may land on the same skill or a different one.\n- **Unrelated to this domain entirely** → treat the previous activation as\n  no longer relevant. Don't carry design-gate guidance into an unrelated\n  research question just because it was active earlier in the session.\n\nThere's no counter or timeout enforcing this — it's a judgment call made\nfresh each turn, the same way you'd judge whether any other piece of earlier\ncontext is still relevant. A hard \"N turns unused → off\" rule isn't something\na hook can enforce anyway: once a skill's content is in the transcript, no\nscript can make the model stop referencing it. Task-continuity judgment is\nthe only lever that actually works.\n\n### Per-category ambiguity preference (`~\u002F.claude\u002Fmasterof\u002Fpreferences.json`)\n\nImplemented — see step 3 above for how each mode is consulted. All six\ncategories default to `{\"mode\": \"always_ask\"}` (unchanged behavior) until the\nuser asks to change one. To change a preference: edit that category's entry\nin `preferences.json` directly (no schema migration needed, it's just a flat\nper-category object) and confirm the change back to the user in plain terms.\n\nFor everything about scanning, classifying new skills, permanent exemptions,\nor generating the status report, see `check-skills`\n(`.\u002Fskills\u002Fcheck-skills\u002FSKILL.md`) — that skill owns all of it now. For\nwhole-pipeline skills where the user must pick exactly one workflow rather\nthan combine several, see `pipelines`\n(`.\u002Fskills\u002Fpipelines\u002FSKILL.md`) — its Activation Protocol variant is\nsingle-select, not multi-select.\n","# master-of — overview\n\nThis plugin bundles:\n- **Domain gates**: `design` \u002F `dev` \u002F `research` \u002F\n  `stock` \u002F `planning` — on-demand activation for a dormant\n  skill in that domain. Each is a thin pointer to the Activation Protocol\n  below.\n- **`check-skills`** — the system's own manager. It scans for\n  new\u002Fremoved skills and plugins (automatically via the `SessionStart` hook,\n  or on demand), classifies new finds into the right gate (or a new one),\n  and reports the current full breakdown read-only. See\n  `.\u002Fskills\u002Fcheck-skills\u002FSKILL.md` for everything about how that works —\n  this file doesn't duplicate it.\n\n## Activation protocol (shared by every domain gate)\n\nEvery domain gate (`design`\u002F`dev`\u002F`research`\u002F`stock`\u002F`planning`) delegates its\nactual behavior here instead of repeating it — each gate's own SKILL.md is\njust a pointer to this section plus its category key and any\ncategory-specific notes.\n\n**When to open a gate at all.** Two ways in, both fine: the user invokes it\n(`\u002Fdesign …`), or you open it yourself because the user asked for actual\n*work* that clearly belongs to that domain (\"이 화면 다듬어줘\", \"경쟁사\n스크래핑해줘\"). Don't open one for a question, an explanation, or chat that\nmerely mentions the topic — opening costs tokens (100-2,800 per gate), so it\nhas to be a task you're about to do. When in doubt, ask in one line rather\nthan opening. Once a gate is open:\n\n1. Read `~\u002F.claude\u002Fmasterof\u002Fgates\u002F\u003Ccategory>.txt` — the pre-rendered index for\n   that one gate (`name | 설명 | Read할 경로`, one line per entry).\n   **Never `Read` `~\u002F.claude\u002Fmasterof\u002Fregistry.json` to activate a gate.**\n   `Read` pulls the whole file, and registry.json is ~60KB \u002F ~16k tokens —\n   more than 2.5x the entire per-session saving this system buys, spent on a\n   single activation. The gate files carry the same name\u002Fdescription\u002Fpath the\n   protocol needs, at 100-2,500 tokens each. registry.json is the source of\n   truth for *classification* (`check-skills` writes it); the `gates\u002F`\n   files are the read path for *activation*, regenerated from it by\n   `scripts\u002Frender-report.ts` on every change.\n2. **Also glance at scale — the related pipelines are already in the file.**\n   Judge whether the request is a narrow, specific task (→ stay with the\n   domain entries) or something project-\u002Fworkflow-scale (\"전체 다 해줘\",\n   \"처음부터 끝까지\", a whole new design direction, a whole phase cycle) that a\n   complete pipeline would handle better. Each gate file ends with a\n   `## 관련 파이프라인` section listing exactly the pipelines whose `domain`\n   is this category, so this costs no extra read. Don't treat `design` and\n   `pipelines` as separate lanes the user has to know to pick between; a\n   design-shaped request should surface a fitting pipeline (e.g. `paint`) as\n   naturally as it surfaces a supporting skill (e.g. `better-ui`), when the\n   scale calls for it. If both a narrow skill and a pipeline plausibly fit,\n   say so and let the user choose which scale they want, rather than guessing.\n3. **Try to match first — don't show anything yet.** Compare what the user\n   actually asked for (the message that triggered this skill, or the task at\n   hand) against each entry's `description`.\n   - One confident match → say so directly (\"이 작업엔 `\u003Cname>`을 쓸게요\") and\n     `Read` its `path`. No list, no round-trip.\n   - A small confident cluster (2-4 entries that clearly belong together for\n     this one task) → name them directly the same way and proceed.\n   - **Genuinely ambiguous** (several entries plausibly fit and none clearly\n     wins) → check `~\u002F.claude\u002Fmasterof\u002Fpreferences.json` for this category\n     before asking:\n     - `{\"mode\": \"always_ask\"}` → print the full list (grouped by `cluster`\n       for `planning`, flat for the others) and ask, offering\n       `전체`\u002F`all`, numbers, names, or (for `planning`) a cluster name.\n     - `{\"mode\": \"fixed_default\", \"default\": \"\u003Cname>\"}` → use that entry\n       without asking, but say which one and that it's the saved default (so\n       the user can override in the moment if it's wrong for this task).\n     - `{\"mode\": \"conditional\", \"rules\": [{\"when\": \"\u003Ccondition, in plain\n       language>\", \"use\": \"\u003Cname>\"}, ...], \"fallback\": \"always_ask\" |\n       \"\u003Cname>\"}` → check each rule's `when` against the task in order (plain\n       judgment, not string matching); use the first match's `use`; if none\n       match, fall back to `fallback` (either ask, or another fixed default).\n     - `{\"mode\": \"smart\"}` → judge from context yourself and ask only when\n       truly tied.\n   - If the user explicitly asks something like \"뭐 있어\" \u002F \"다 보여줘\" \u002F \"전체\n     목록\", always show the full list regardless of preference mode — that's\n     an explicit request to see everything, not an ambiguous match.\n4. `Read` only what got selected — never the whole category \"just in case.\"\n   Bulk-reading every entry in a category on every invocation is exactly the\n   always-on cost this system exists to avoid; a gate that does that is a bug,\n   not a feature.\n   **Carry the base directory with it.** A gated skill is activated by\n   `Read`ing its `SKILL.md` directly, which — unlike a real `Skill` invocation\n   — supplies no \"Base directory for this skill\" line. Many of these skills\n   ship companion files and refer to them relatively (`RECIPES.md`,\n   `STANDARDS.md`, `reference\u002F`, `scripts\u002F…`); with no anchor those references\n   are dead ends. So treat the directory containing the `SKILL.md` you just\n   read (the path in the gate index, minus the filename) as that skill's base\n   directory, and resolve every relative path in its body against it. State it\n   once when you activate (\"base: `…\u002Fskills-library\u002Fdesign\u002Fanimate\u002F`\") so it's\n   explicit rather than re-derived each time you follow a reference.\n   **Reading a skill is not the same as invoking it — three things the\n   `Skill` tool would have done for you that you now do yourself:**\n   - `$ARGUMENTS` is not substituted. Many skills (every GSD skill, for\n     one) say things like \"Phase number: $ARGUMENTS\". Treat whatever the\n     user gave after the skill's name — or the specific thing they asked for\n     — as that value, and if the skill needs an argument the user didn't\n     supply, ask for it the way the skill itself says to (most say what to\n     do when it's omitted).\n   - `allowed-tools` in the frontmatter doesn't pre-approve anything. The\n     skill will work; you may just get permission prompts it wouldn't\n     normally trigger. Don't read that as the skill being broken.\n   - `disable-model-invocation: true` in the frontmatter means the author\n     wants it run only when the user explicitly asks for it BY NAME — never\n     on your own judgment. Opening a gate yourself (the auto-open rule) does\n     not lift that: if the best match carries this flag and the user didn't\n     name it, mention it as an option and let them choose; don't activate it.\n   **Commands and agents are gated the same way, and the index tags them.**\n   Plugins ship more than skills, and every command and agent costs\n   always-on tokens exactly like a skill does, so they're classified into\n   the same gates. An index line starting `[커맨드]` or `[에이전트]` is\n   activated differently from a skill:\n   - `[커맨드]` → `Read` it. The body is the prompt a `\u002Fname args` invocation\n     would have run; `$ARGUMENTS` is the user's request (same rule as\n     above). Follow it as those instructions, base directory rule included.\n   - `[에이전트]` → `Read` it. The frontmatter's `description`\u002F`tools`\u002F\n     `model` and the body (its system prompt) are what the harness would\n     have used to register it as a `subagent_type`. Since it isn't\n     registered, spawn it yourself: `Agent` with `subagent_type:\n     \"general-purpose\"`, `prompt` = the body followed by the concrete task,\n     `model` = the frontmatter's if it names one you can pass, and the\n     `tools` line restated inside the prompt as a constraint (\"only use\n     …\"). Say once that you're running it this way.\n   - **A gated skill that says \"spawn `subagent_type: X`\"** where X isn't in\n     the available agent types is *not* broken: check the current gate's\n     index (then `_all.txt`) for `[에이전트] X` and run it as above. This is\n     the normal path for frameworks whose skills and agents were gated\n     together. It costs one `Read` of X's file per spawn on your side — a\n     framework that spawns many agents per workflow pays that each time,\n     which is why `check-skills` asks before gating a large agent set.\n   **Honor `[의존: …]` notes.** An index line can carry\n   `[의존: \u003Cplugin> \u003Ccomponent> — 현재 꺼짐, 없으면 동작 불가]`. That means the\n   skill's plugin is disabled, so the MCP server \u002F agents it needs don't\n   exist this session. If it says 없으면 동작 불가, don't activate it — tell\n   the user it can't work right now and what would fix it (`claude plugin\n   enable \u003Cplugin>`, effective next session), and offer the closest working\n   alternative instead. Without that suffix (a partial), activate but say\n   which parts won't run. The status is as of the last render; the session-\n   start hook is the fresh source and will already have said so if it's off.\n5. **A match is a candidate, not an obligation.** The index answers \"what\n   exists in this domain\", which is not the same question as \"does this task\n   actually need it\". If the best match is only topically adjacent — it shares\n   the subject but not the job (a \"how to build X\" guide when the task is\n   \"decide whether to build X\"; a reviewer skill when the task is to gather\n   facts) — say so and skip the Read instead of activating it for form's sake.\n   Reading a 2-3k token skill that doesn't change what you do costs more than\n   the gate ever saved.\n6. Briefly confirm what's active, then proceed with the actual task.\n\nScope: one-shot, in-context only. Nothing persists, nothing is written to\nsettings. Re-invoke the gate for the next unrelated task in that domain.\n\n### Multi-activation — several skills, and several domains, at once\n\nActivation is **not** limited to one skill. A real request often needs two or\nthree things working together (\"이 화면 리뷰하고 접근성까지 봐줘\", \"경쟁사\n스크래핑해서 그 디자인 분석해줘\"), and forcing the user to invoke gates one at a\ntime, serially, is the single most annoying way this system can behave.\n\n**Within one domain** — step 3 already covers it: a confident cluster of 2-4\nentries is named and read together, not one at a time.\n\n**Across domains** — same idea, one level up:\n\n- If you can name the domains from the request (usually you can — \"스크래핑\"\n  → `research`, \"디자인 분석\" → `design`), read just those gate files. Two\n  gate files is typically 1-2k tokens total, far cheaper than the combined\n  index.\n- Only when the domain set is genuinely unclear, read\n  `~\u002F.claude\u002Fmasterof\u002Fgates\u002F_all.txt` — every category in one file (~6k\n  tokens). It's the cross-domain fallback, not the default.\n- Then match within the combined set and activate what fits, naming each one\n  and which domain it came from (\"`firecrawl-scrape`(research) + `better-ui`\n  (design) 두 개 붙일게요\"). Read only those.\n- **Cap it at ~4 skills.** Beyond that the activations start crowding out the\n  actual task, and it's a sign the request should be split — or that a single\n  `pipelines` entry is the right answer instead of five supporting skills.\n- The `pipelines` single-select rule still holds across domains: at most ONE\n  pipeline, and a pipeline generally shouldn't be combined with supporting\n  skills from other domains — it already runs its own end-to-end workflow.\n\nInvoking the root `master-of` skill directly (`\u002Fmaster-of \u003C요청>`) is the\nexplicit cross-domain entry point: it runs exactly this step — match across\nevery category first, then activate the 1-4 skills that fit, from however\nmany domains they happen to live in.\n\n### Staying activated across a continued task\n\nDon't re-run the full matching step from scratch on every single message.\nWhen a new user request comes in while something is already active from this\ndomain:\n- **Same or clearly continuing task** → keep using what's already active, no\n  re-invocation needed. Do a quick silent check whether a different entry in\n  the category would now fit better (the task may have shifted); if so,\n  mention it as a one-line suggestion rather than silently switching.\n- **A new, different task in the same domain** → re-run the matching step\n  fresh (steps 1-4 above); it may land on the same skill or a different one.\n- **Unrelated to this domain entirely** → treat the previous activation as\n  no longer relevant. Don't carry design-gate guidance into an unrelated\n  research question just because it was active earlier in the session.\n\nThere's no counter or timeout enforcing this — it's a judgment call made\nfresh each turn, the same way you'd judge whether any other piece of earlier\ncontext is still relevant. A hard \"N turns unused → off\" rule isn't something\na hook can enforce anyway: once a skill's content is in the transcript, no\nscript can make the model stop referencing it. Task-continuity judgment is\nthe only lever that actually works.\n\n### Per-category ambiguity preference (`~\u002F.claude\u002Fmasterof\u002Fpreferences.json`)\n\nImplemented — see step 3 above for how each mode is consulted. All six\ncategories default to `{\"mode\": \"always_ask\"}` (unchanged behavior) until the\nuser asks to change one. To change a preference: edit that category's entry\nin `preferences.json` directly (no schema migration needed, it's just a flat\nper-category object) and confirm the change back to the user in plain terms.\n\nFor everything about scanning, classifying new skills, permanent exemptions,\nor generating the status report, see `check-skills`\n(`.\u002Fskills\u002Fcheck-skills\u002FSKILL.md`) — that skill owns all of it now. For\nwhole-pipeline skills where the user must pick exactly one workflow rather\nthan combine several, see `pipelines`\n(`.\u002Fskills\u002Fpipelines\u002FSKILL.md`) — its Activation Protocol variant is\nsingle-select, not multi-select.\n",{"name":5,"description":7,"allowed-tools":23},"Read","main","89da93a383f2635c66566dac6a702d9a87560b1c","plugins\u002Fmaster-of",0,"https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of","https:\u002F\u002Fgithub.com\u002Fdevrudals","2026-11-02 17:03:29",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add devrudals\u002Fmaster-of --skill master-of",{"tool":37,"command":38},"bunx","bunx skills add devrudals\u002Fmaster-of --skill master-of",{"tool":40,"command":41},"pnpm","pnpm dlx skills add devrudals\u002Fmaster-of --skill master-of","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002FSKILL.md",true,[45,51,57,61],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},".claude-plugin\u002Fplugin.json","REFERENCE","参考文档",1724,"https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002F.claude-plugin\u002Fplugin.json",{"path":52,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":55,"sourceHtmlUrl":56},"bin\u002Fmo.mjs","SCRIPT","脚本",127121,"https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002Fbin\u002Fmo.mjs",{"path":58,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":59,"sourceHtmlUrl":60},"hooks\u002Fhooks.json",250,"https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002Fhooks\u002Fhooks.json",{"path":62,"resourceType":53,"resourceTypeDesc":54,"sizeBytes":63,"sourceHtmlUrl":64},"hooks\u002Fsession-start.sh",1075,"https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fplugins\u002Fmaster-of\u002Fhooks\u002Fsession-start.sh",[66,73,80,87,94,101],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M7CK032J522RKKR1BMC21191","flutter-preview","Flutter 위젯 인터랙티브 프리뷰 시스템 및 반응형 UI 레이아웃","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-gemini\u002Fskills\u002Fflutter-preview\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":18},"01M7CK00W1R6R4AVZ835EFXSP3","dart-unit-test","Dart 및 Flutter 단위 테스트 작성 및 코드 커버리지 리포트 생성","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-gemini\u002Fskills\u002Fdart-unit-test\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M7CJZZCQDKQS0YKWJ7125AGK","web-scraper","웹 브라우징 및 동적 페이지 데이터 크롤링, 텍스트 정제","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-claude\u002Fskills\u002Fweb-scraper\u002FSKILL.md",[],[],{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"publishedAt":18},"01M7CJZVQ22TT4BH4YFSAM1DAN","sprint-planner","GSD 스프린트 백로그 구성, 태스크 분해 및 마일스톤 추적","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-claude\u002Fskills\u002Fsprint-planner\u002FSKILL.md",[],[],{"skillNo":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"publishedAt":18},"01M7CJZSZZPW678EKDP67MA1Q9","krx-analyzer","한국 거래소(KRX) 상장 기업 재무제표 및 공시 데이터 분석","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-claude\u002Fskills\u002Fkrx-analyzer\u002FSKILL.md",[],[],{"skillNo":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"publishedAt":18},"01M7CJZPYP5AHANX9CFXER0TP9","git-workflow","Git 브랜치 전략, 커밋 메시지 컨벤션, PR 템플릿 자동화","https:\u002F\u002Fgithub.com\u002Fdevrudals\u002Fmaster-of\u002Fblob\u002Fmain\u002Fsandbox\u002Fmock-claude\u002Fskills\u002Fgit-workflow\u002FSKILL.md",[],[]]