[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DDFNCEJQG9SVSAR2FG7QM4":3,"skill-same:01M7DDFNCEJQG9SVSAR2FG7QM4":49},{"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":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":32,"ownerDescription":33,"updatedAt":34,"install":35,"resources":48},"01M7DDFNCEJQG9SVSAR2FG7QM4","explore",null,"Explores a codebase to build a complete, evidence-backed model of the project — architecture, module boundaries, entry points, data flow, conventions, standards, tooling, and its documented preferences — then returns a structured, summarized report for context gaining and planning. Use when the user types \u002Fexplore, asks to explore or understand the codebase, wants a project overview, needs onboarding, or is starting a session and wants context before planning work. Read-only by default.","Marvellbrazil\u002Fmy-agentic-skills","Marvellbrazil","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184349613?v=4","01M7DCJPCRHP7SZS0T5H1GJZ8F","Marvello Faisal","01M7DCJPDCZDB94VG0QXGC6RC3",5,"MIT","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fexplore\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: explore\ndescription: Explores a codebase to build a complete, evidence-backed model of the project — architecture, module boundaries, entry points, data flow, conventions, standards, tooling, and its documented preferences — then returns a structured, summarized report for context gaining and planning. Use when the user types \u002Fexplore, asks to explore or understand the codebase, wants a project overview, needs onboarding, or is starting a session and wants context before planning work. Read-only by default.\nallowed-tools: Read Grep Glob Bash\n---\n\n# Explore\n\nBuild an accurate mental model of a codebase and hand it back in a form the user can act on.\n\nThis is a **read-only** skill unless the user asks for a written artifact. It answers three questions: what is this project, how is it organized, and what rules govern changes to it. The output is the input to planning.\n\n## Principles\n\n1. **Evidence, not impression.** Every claim names the file it came from. \"Uses Postgres\" is weak; \"Uses Postgres — `docker-compose.yml:14`, `prisma\u002Fschema.prisma:1`\" is usable.\n2. **Conventions before code.** The documented rules (`AGENTS.md`, `CONTRIBUTING.md`, lint configs) constrain every future change. Read them first; they are cheap and they are authoritative.\n3. **Read the manifest, not the whole tree.** A manifest declares the project's intent. Source files confirm it. Do not read 400 files to answer a question a manifest answers.\n4. **Prefer structure over content.** Directory shape, file names, and export signatures carry most of the signal at a fraction of the cost.\n5. **Say what you did not determine.** An honest gap is useful. A confident guess about the architecture is actively harmful, because the user will plan on top of it.\n6. **Do not modify anything.** Exploration must not leave a trace.\n\n## Phase 0 — Scope\n\nDefault to a full pass. Narrow it when the user named a target or the repo is large enough that a full pass would be wasteful.\n\n| Signal                        | Action                                                                          |\n| ----------------------------- | ------------------------------------------------------------------------------- |\n| Bare `\u002Fexplore`               | Full pass                                                                       |\n| `\u002Fexplore the payment module` | Narrow to that subtree, plus its direct dependencies and consumers              |\n| Monorepo with 20 packages     | Ask which package, or explore the workspace root and one representative package |\n| More than ~5,000 source files | Ask for a target; do not attempt an exhaustive read                             |\n\nIf you must narrow, say so and name what you skipped. Never silently truncate coverage — the user will assume the report is complete.\n\n## Phase 1 — Identity and Intent\n\nAnswer \"what is this and what is it for\" before anything else.\n\n```bash\nls -a\ncat README.md 2>\u002Fdev\u002Fnull | head -80\n```\n\nRead the manifest. It declares the runtime, dependencies, and the project's own entry points.\n\n```bash\ncat package.json 2>\u002Fdev\u002Fnull\ncat pyproject.toml 2>\u002Fdev\u002Fnull\ncat composer.json Gemfile go.mod Cargo.toml pom.xml build.gradle.kts 2>\u002Fdev\u002Fnull\ncat deno.json bun.lockb 2>\u002Fdev\u002Fnull\n```\n\nExtract and record:\n\n| Field                             | Where to look                                                                                                                      |\n| --------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |\n| Project name and one-line purpose | `README.md`, manifest `name`\u002F`description`                                                                                         |\n| Runtime and version               | `engines`, `requires-python`, `go`, `toolchain`, `.nvmrc`, `.python-version`                                                       |\n| Package manager                   | `packageManager`, the lockfile present (`bun.lockb`, `pnpm-lock.yaml`, `yarn.lock`, `package-lock.json`, `uv.lock`, `poetry.lock`) |\n| Primary language and framework    | Dependency list                                                                                                                    |\n| Available scripts                 | `scripts` in the manifest — these are the project's real verbs                                                                     |\n| Declared entry points             | `main`, `module`, `exports`, `bin`, `[project.scripts]`                                                                            |\n\n**The scripts block is the highest-value field in the manifest.** It is the project's vocabulary for build, test, lint, and run. Record it verbatim.\n\n## Phase 2 — Conventions and Standards\n\nDo this before reading source. The rules constrain how you interpret everything else.\n\n```bash\nfor f in AGENTS.md CLAUDE.md CODEBASE.md CONTRIBUTING.md ARCHITECTURE.md STYLE.md CONVENTIONS.md SECURITY.md; do\n  [ -f \"$f\" ] && echo \"=== $f ===\" && cat \"$f\"\ndone\nfind . -maxdepth 3 -iname '*.md' -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' \\\n  -not -path '*\u002Fvendor\u002F*' 2>\u002Fdev\u002Fnull | head -40\n```\n\n```bash\nls -a | grep -E '^\\.(eslint|prettier|editorconfig|stylelint|rubocop|golangci|swiftlint|clang-format)'\ncat .editorconfig eslint.config.js eslint.config.mjs .eslintrc* .prettierrc* 2>\u002Fdev\u002Fnull\ncat pyproject.toml 2>\u002Fdev\u002Fnull | sed -n '\u002F\\[tool\\.\u002F,\u002F^\\[\u002Fp'\ncat .golangci.yml .rubocop.yml phpstan.neon analysis_options.yaml 2>\u002Fdev\u002Fnull\n```\n\nRecord a conventions table. This is what the user most needs and least often has written down:\n\n| Concern           | Rule                                          | Source             | Enforced?   |\n| ----------------- | --------------------------------------------- | ------------------ | ----------- |\n| Language standard | PHP 8.3, `declare(strict_types=1)`            | `AGENTS.md`        | No          |\n| Formatting        | 2-space, single quotes, no semicolons         | `.prettierrc`      | Yes (CI)    |\n| Type strictness   | TS `strict: true`, `noUncheckedIndexedAccess` | `tsconfig.json`    | Yes (build) |\n| Naming            | kebab-case files, PascalCase components       | `CONTRIBUTING.md`  | No          |\n| Import style      | Absolute from `@\u002F`, no barrels                | `eslint.config.js` | Yes (lint)  |\n| Comments          | No inline comments; self-documenting code     | `AGENTS.md`        | No          |\n| Commit style      | Conventional Commits, no co-author            | `AGENTS.md`        | No          |\n| Testing           | Vitest, colocated `*.test.ts`                 | `package.json`     | Yes (CI)    |\n\n**Mark whether each rule is enforced.** A rule enforced by CI is a fact; a rule in a prose document is an intention. This distinction tells the user where drift is likely.\n\nIf there is **no** conventions document, say so explicitly and derive the conventions from the code instead, labeling them as observed rather than declared.\n\n## Phase 3 — Structure and Boundaries\n\nMap the tree without reading every file.\n\n```bash\nfind . -maxdepth 2 -type d -not -path '.\u002F.git*' -not -path '*\u002Fnode_modules*' \\\n  -not -path '*\u002Fvendor*' -not -path '*\u002Fdist*' -not -path '*\u002Fbuild*' \\\n  -not -path '*\u002F.next*' -not -path '*\u002F__pycache__*' | sort | head -60\n```\n\n```bash\nfind . -type f -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' -not -path '*\u002Fvendor\u002F*' \\\n  -not -path '*\u002Fdist\u002F*' -not -path '*\u002Fbuild\u002F*' \\\n  | sed 's\u002F.*\\.\u002F\u002F' | sort | uniq -c | sort -rn | head -20\n```\n\n```bash\nfind . -maxdepth 3 -type f \\( -name 'index.*' -o -name 'main.*' -o -name 'app.*' \\\n  -o -name 'server.*' -o -name 'routes.*' -o -name 'mod.rs' -o -name '__init__.py' \\) \\\n  -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' -not -path '*\u002Fvendor\u002F*' 2>\u002Fdev\u002Fnull | head -30\n```\n\nIdentify and label:\n\n- **Architecture style.** Layer-first (`controllers\u002F`, `services\u002F`, `models\u002F`) vs feature-first (`orders\u002F`, `billing\u002F`) vs hexagonal vs monorepo packages. Name it from the evidence.\n- **Module boundaries.** Which directories are cohesive units with a public surface, and which are shared utilities.\n- **Entry points.** Where execution begins, per target: HTTP server, CLI, worker, scheduled job, mobile app, library export.\n- **The dependency direction.** Which layers may import which. A violation here is the most common architectural defect; note any you observe.\n\nDetect boundaries from import edges, not from directory names:\n\n```bash\ngrep -rn --include='*.ts' --include='*.tsx' -E \"^import .* from ['\\\"]\\.\\.\" src\u002F 2>\u002Fdev\u002Fnull | head -20\n```\n\n## Phase 4 — Data and Runtime\n\n```bash\ncat docker-compose.yml docker-compose.yaml Dockerfile .dockerignore 2>\u002Fdev\u002Fnull | head -100\nfind . -maxdepth 3 \\( -name '*.sql' -o -name 'schema.prisma' -o -name '*.migration' -o -path '*migrations*' \\) \\\n  -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' 2>\u002Fdev\u002Fnull | head -30\nfind . -maxdepth 2 -name '.env*' -not -path '.\u002F.git\u002F*' 2>\u002Fdev\u002Fnull\n```\n\nRecord: datastore and version, cache, queue, external services, and how configuration is supplied. **Never print the contents of a `.env` file.** List which keys it declares, and treat values as secrets:\n\n```bash\nsed 's\u002F=.*\u002F=\u003Credacted>\u002F' .env.example 2>\u002Fdev\u002Fnull | head -30\n```\n\n## Phase 5 — Workflow and Tooling\n\n```bash\ncat Makefile justfile Taskfile.yml 2>\u002Fdev\u002Fnull | head -60\nls -R .github\u002Fworkflows 2>\u002Fdev\u002Fnull\ncat .github\u002Fworkflows\u002F*.yml 2>\u002Fdev\u002Fnull | grep -E 'name:|run:|uses:' | head -40\n```\n\nRecord: the build command, the test command, the lint command, the dev-server command, and what CI actually gates on. **CI is the real definition of \"done\" in the repo** — if CI runs `lint && test && build`, that is the contract.\n\nNote any pre-commit hooks and what they enforce.\n\n## Phase 6 — Synthesize\n\nRead selectively, only to confirm or refute what the structure implies. Read one representative file per architectural layer — not all of them.\n\n```bash\nwc -l $(git ls-files '*.ts' '*.tsx' 2>\u002Fdev\u002Fnull | head -50) 2>\u002Fdev\u002Fnull | sort -rn | head -15\n```\n\nThe largest files are where the architecture is weakest and where the user's future work will hurt. Report them.\n\n```bash\ngit log --oneline -15\ngit log --format='%s' -60 | sed 's\u002F(.*\u002F\u002F' | sed 's\u002F:.*\u002F\u002F' | sort | uniq -c | sort -rn\n```\n\nThe commit history reveals what the team actually works on and whether they follow their own commit convention.\n\n## Output\n\nDeliver a single structured report. Lead with the summary — the user may read only that.\n\n```markdown\n# Project Profile — \u003Cname>\n\n\u003C2-3 sentence summary: what it is, what it does, how it is built.>\n\n**Stack** TypeScript 5.6 · React 19 · Vite 6 · Postgres 16 · Bun\n**Architecture** Feature-first, colocated tests, no barrel files\n**Entry points** `src\u002Fmain.tsx` (browser), `server\u002Findex.ts` (API)\n**Health** 1 conventions gap, 3 oversized files, no type-check in CI\n\n## 1. What This Is\n\n\u003Cpurpose, domain, who uses it, and what it is not>\n\n## 2. Stack and Runtime\n\n| Layer | Choice | Version | Evidence |\n| ----- | ------ | ------- | -------- |\n\n## 3. Structure\n\n\u003Cannotated tree of the meaningful directories, with a one-line purpose each>\n\n## 4. Module Boundaries\n\n\u003Cwhich units exist, their public surface, and the allowed dependency direction>\n\n## 5. Entry Points and Flow\n\n\u003Chow a request or command travels from entry to storage, naming the files>\n\n## 6. Conventions and Standards\n\n\u003Cthe conventions table from Phase 2, including what is enforced vs merely stated>\n\n## 7. Commands\n\n| Purpose | Command     | Source          |\n| ------- | ----------- | --------------- |\n| Dev     | `bun dev`   | package.json:12 |\n| Test    | `bun test`  | package.json:14 |\n| Lint    | `bun lint`  | package.json:15 |\n| Build   | `bun build` | package.json:16 |\n\n## 8. CI Gates\n\n\u003Cwhat CI enforces, and what it does not>\n\n## 9. Observations\n\n\u003Cthe non-obvious: architectural risks, oversized files, drift from documented\nrules, missing tests, dead code, inconsistencies worth a \u002Fs13n pass>\n\n## 10. Unknowns\n\n\u003Cwhat could not be determined, and the file or question that would resolve it>\n```\n\n## Output Rules\n\n- **Summarize; never dump.** Do not paste file contents or whole manifests. Extract and interpret.\n- **Every claim carries evidence.** A file path, a line number, or a command output.\n- **Separate declared from observed.** \"Documented: no inline comments\" vs \"Observed: 3 files use inline comments.\"\n- **Rank observations by what the user should do first.** A missing CI type-check outranks a stale comment.\n- **Keep it readable.** Tables over prose, a tree over a paragraph, and no wall of text. If a section has nothing, omit it rather than printing \"N\u002FA\".\n- **Flag genuine uncertainty as uncertainty.** Write \"could not determine\" and move on.\n- **Do not recommend a rewrite.** Describe the project as it is. Recommendations belong in a planning step the user asks for.\n\n## Persisting the Profile\n\nIf the user wants the model kept for later sessions, write it to `.claude\u002Fcontext\u002Fproject-profile.md` or a path they name, and tell them the path. Do not create a file unless asked — exploration must not leave a trace by default.\n\n## Depth\n\nScale the effort to the repository and the ask.\n\n| Repository                     | Approach                                                                                     |\n| ------------------------------ | -------------------------------------------------------------------------------------------- |\n| Small (under ~50 source files) | Read the manifests, the conventions docs, and most source files. Be exhaustive.              |\n| Medium (~50-500 files)         | Manifests, conventions, structure, one representative file per layer, and the largest files. |\n| Large (~500-5,000)             | Manifests, conventions, structure, and CI only. Ask the user to name a target for depth.     |\n| Monorepo                       | Workspace root plus the packages the user names, or one representative package.              |\n\n## Common Mistakes\n\n| Mistake                                                       | Why It Breaks                                                     | Correct Approach                                            |\n| ------------------------------------------------------------- | ----------------------------------------------------------------- | ----------------------------------------------------------- |\n| Skipping the conventions docs                                 | You describe a project and miss the rules that govern changing it | Read `AGENTS.md`, `CONTRIBUTING.md`, and lint configs first |\n| Reading files instead of structure                            | Burns enormous context for a fraction of the signal               | Read manifests, tree shape, and signatures                  |\n| Reporting a convention as enforced when it is only documented | The user plans on a gate that does not exist                      | Check whether CI or lint actually enforces it               |\n| Guessing the architecture from directory names                | Names lie; `services\u002F` is often not a service layer               | Confirm with import edges and one representative file       |\n| Claiming completeness after sampling                          | The user plans on a partial picture they believe is whole         | State what you skipped, explicitly                          |\n| Dumping file contents                                         | Unreadable and destroys the value of the summary                  | Extract, interpret, cite                                    |\n| Printing a `.env` file                                        | Leaks secrets into the transcript and the user's history          | List keys only; redact values                               |\n| Writing files unprompted                                      | Exploration should leave no trace                                 | Read-only unless asked                                      |\n| Recommending a rewrite                                        | Not what was asked, and it undermines trust in the findings       | Describe; let the user decide                               |\n| Reporting stale docs as fact                                  | Docs drift; the code is the truth                                 | Separate declared from observed, and note disagreement      |\n| One giant prose section                                       | Unreadable                                                        | Tables, trees, and a scannable structure                    |\n| Ignoring the commit history                                   | It reveals real priorities and convention drift                   | Read the recent log and the subject patterns                |\n","# Explore\n\nBuild an accurate mental model of a codebase and hand it back in a form the user can act on.\n\nThis is a **read-only** skill unless the user asks for a written artifact. It answers three questions: what is this project, how is it organized, and what rules govern changes to it. The output is the input to planning.\n\n## Principles\n\n1. **Evidence, not impression.** Every claim names the file it came from. \"Uses Postgres\" is weak; \"Uses Postgres — `docker-compose.yml:14`, `prisma\u002Fschema.prisma:1`\" is usable.\n2. **Conventions before code.** The documented rules (`AGENTS.md`, `CONTRIBUTING.md`, lint configs) constrain every future change. Read them first; they are cheap and they are authoritative.\n3. **Read the manifest, not the whole tree.** A manifest declares the project's intent. Source files confirm it. Do not read 400 files to answer a question a manifest answers.\n4. **Prefer structure over content.** Directory shape, file names, and export signatures carry most of the signal at a fraction of the cost.\n5. **Say what you did not determine.** An honest gap is useful. A confident guess about the architecture is actively harmful, because the user will plan on top of it.\n6. **Do not modify anything.** Exploration must not leave a trace.\n\n## Phase 0 — Scope\n\nDefault to a full pass. Narrow it when the user named a target or the repo is large enough that a full pass would be wasteful.\n\n| Signal                        | Action                                                                          |\n| ----------------------------- | ------------------------------------------------------------------------------- |\n| Bare `\u002Fexplore`               | Full pass                                                                       |\n| `\u002Fexplore the payment module` | Narrow to that subtree, plus its direct dependencies and consumers              |\n| Monorepo with 20 packages     | Ask which package, or explore the workspace root and one representative package |\n| More than ~5,000 source files | Ask for a target; do not attempt an exhaustive read                             |\n\nIf you must narrow, say so and name what you skipped. Never silently truncate coverage — the user will assume the report is complete.\n\n## Phase 1 — Identity and Intent\n\nAnswer \"what is this and what is it for\" before anything else.\n\n```bash\nls -a\ncat README.md 2>\u002Fdev\u002Fnull | head -80\n```\n\nRead the manifest. It declares the runtime, dependencies, and the project's own entry points.\n\n```bash\ncat package.json 2>\u002Fdev\u002Fnull\ncat pyproject.toml 2>\u002Fdev\u002Fnull\ncat composer.json Gemfile go.mod Cargo.toml pom.xml build.gradle.kts 2>\u002Fdev\u002Fnull\ncat deno.json bun.lockb 2>\u002Fdev\u002Fnull\n```\n\nExtract and record:\n\n| Field                             | Where to look                                                                                                                      |\n| --------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |\n| Project name and one-line purpose | `README.md`, manifest `name`\u002F`description`                                                                                         |\n| Runtime and version               | `engines`, `requires-python`, `go`, `toolchain`, `.nvmrc`, `.python-version`                                                       |\n| Package manager                   | `packageManager`, the lockfile present (`bun.lockb`, `pnpm-lock.yaml`, `yarn.lock`, `package-lock.json`, `uv.lock`, `poetry.lock`) |\n| Primary language and framework    | Dependency list                                                                                                                    |\n| Available scripts                 | `scripts` in the manifest — these are the project's real verbs                                                                     |\n| Declared entry points             | `main`, `module`, `exports`, `bin`, `[project.scripts]`                                                                            |\n\n**The scripts block is the highest-value field in the manifest.** It is the project's vocabulary for build, test, lint, and run. Record it verbatim.\n\n## Phase 2 — Conventions and Standards\n\nDo this before reading source. The rules constrain how you interpret everything else.\n\n```bash\nfor f in AGENTS.md CLAUDE.md CODEBASE.md CONTRIBUTING.md ARCHITECTURE.md STYLE.md CONVENTIONS.md SECURITY.md; do\n  [ -f \"$f\" ] && echo \"=== $f ===\" && cat \"$f\"\ndone\nfind . -maxdepth 3 -iname '*.md' -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' \\\n  -not -path '*\u002Fvendor\u002F*' 2>\u002Fdev\u002Fnull | head -40\n```\n\n```bash\nls -a | grep -E '^\\.(eslint|prettier|editorconfig|stylelint|rubocop|golangci|swiftlint|clang-format)'\ncat .editorconfig eslint.config.js eslint.config.mjs .eslintrc* .prettierrc* 2>\u002Fdev\u002Fnull\ncat pyproject.toml 2>\u002Fdev\u002Fnull | sed -n '\u002F\\[tool\\.\u002F,\u002F^\\[\u002Fp'\ncat .golangci.yml .rubocop.yml phpstan.neon analysis_options.yaml 2>\u002Fdev\u002Fnull\n```\n\nRecord a conventions table. This is what the user most needs and least often has written down:\n\n| Concern           | Rule                                          | Source             | Enforced?   |\n| ----------------- | --------------------------------------------- | ------------------ | ----------- |\n| Language standard | PHP 8.3, `declare(strict_types=1)`            | `AGENTS.md`        | No          |\n| Formatting        | 2-space, single quotes, no semicolons         | `.prettierrc`      | Yes (CI)    |\n| Type strictness   | TS `strict: true`, `noUncheckedIndexedAccess` | `tsconfig.json`    | Yes (build) |\n| Naming            | kebab-case files, PascalCase components       | `CONTRIBUTING.md`  | No          |\n| Import style      | Absolute from `@\u002F`, no barrels                | `eslint.config.js` | Yes (lint)  |\n| Comments          | No inline comments; self-documenting code     | `AGENTS.md`        | No          |\n| Commit style      | Conventional Commits, no co-author            | `AGENTS.md`        | No          |\n| Testing           | Vitest, colocated `*.test.ts`                 | `package.json`     | Yes (CI)    |\n\n**Mark whether each rule is enforced.** A rule enforced by CI is a fact; a rule in a prose document is an intention. This distinction tells the user where drift is likely.\n\nIf there is **no** conventions document, say so explicitly and derive the conventions from the code instead, labeling them as observed rather than declared.\n\n## Phase 3 — Structure and Boundaries\n\nMap the tree without reading every file.\n\n```bash\nfind . -maxdepth 2 -type d -not -path '.\u002F.git*' -not -path '*\u002Fnode_modules*' \\\n  -not -path '*\u002Fvendor*' -not -path '*\u002Fdist*' -not -path '*\u002Fbuild*' \\\n  -not -path '*\u002F.next*' -not -path '*\u002F__pycache__*' | sort | head -60\n```\n\n```bash\nfind . -type f -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' -not -path '*\u002Fvendor\u002F*' \\\n  -not -path '*\u002Fdist\u002F*' -not -path '*\u002Fbuild\u002F*' \\\n  | sed 's\u002F.*\\.\u002F\u002F' | sort | uniq -c | sort -rn | head -20\n```\n\n```bash\nfind . -maxdepth 3 -type f \\( -name 'index.*' -o -name 'main.*' -o -name 'app.*' \\\n  -o -name 'server.*' -o -name 'routes.*' -o -name 'mod.rs' -o -name '__init__.py' \\) \\\n  -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' -not -path '*\u002Fvendor\u002F*' 2>\u002Fdev\u002Fnull | head -30\n```\n\nIdentify and label:\n\n- **Architecture style.** Layer-first (`controllers\u002F`, `services\u002F`, `models\u002F`) vs feature-first (`orders\u002F`, `billing\u002F`) vs hexagonal vs monorepo packages. Name it from the evidence.\n- **Module boundaries.** Which directories are cohesive units with a public surface, and which are shared utilities.\n- **Entry points.** Where execution begins, per target: HTTP server, CLI, worker, scheduled job, mobile app, library export.\n- **The dependency direction.** Which layers may import which. A violation here is the most common architectural defect; note any you observe.\n\nDetect boundaries from import edges, not from directory names:\n\n```bash\ngrep -rn --include='*.ts' --include='*.tsx' -E \"^import .* from ['\\\"]\\.\\.\" src\u002F 2>\u002Fdev\u002Fnull | head -20\n```\n\n## Phase 4 — Data and Runtime\n\n```bash\ncat docker-compose.yml docker-compose.yaml Dockerfile .dockerignore 2>\u002Fdev\u002Fnull | head -100\nfind . -maxdepth 3 \\( -name '*.sql' -o -name 'schema.prisma' -o -name '*.migration' -o -path '*migrations*' \\) \\\n  -not -path '.\u002F.git\u002F*' -not -path '*\u002Fnode_modules\u002F*' 2>\u002Fdev\u002Fnull | head -30\nfind . -maxdepth 2 -name '.env*' -not -path '.\u002F.git\u002F*' 2>\u002Fdev\u002Fnull\n```\n\nRecord: datastore and version, cache, queue, external services, and how configuration is supplied. **Never print the contents of a `.env` file.** List which keys it declares, and treat values as secrets:\n\n```bash\nsed 's\u002F=.*\u002F=\u003Credacted>\u002F' .env.example 2>\u002Fdev\u002Fnull | head -30\n```\n\n## Phase 5 — Workflow and Tooling\n\n```bash\ncat Makefile justfile Taskfile.yml 2>\u002Fdev\u002Fnull | head -60\nls -R .github\u002Fworkflows 2>\u002Fdev\u002Fnull\ncat .github\u002Fworkflows\u002F*.yml 2>\u002Fdev\u002Fnull | grep -E 'name:|run:|uses:' | head -40\n```\n\nRecord: the build command, the test command, the lint command, the dev-server command, and what CI actually gates on. **CI is the real definition of \"done\" in the repo** — if CI runs `lint && test && build`, that is the contract.\n\nNote any pre-commit hooks and what they enforce.\n\n## Phase 6 — Synthesize\n\nRead selectively, only to confirm or refute what the structure implies. Read one representative file per architectural layer — not all of them.\n\n```bash\nwc -l $(git ls-files '*.ts' '*.tsx' 2>\u002Fdev\u002Fnull | head -50) 2>\u002Fdev\u002Fnull | sort -rn | head -15\n```\n\nThe largest files are where the architecture is weakest and where the user's future work will hurt. Report them.\n\n```bash\ngit log --oneline -15\ngit log --format='%s' -60 | sed 's\u002F(.*\u002F\u002F' | sed 's\u002F:.*\u002F\u002F' | sort | uniq -c | sort -rn\n```\n\nThe commit history reveals what the team actually works on and whether they follow their own commit convention.\n\n## Output\n\nDeliver a single structured report. Lead with the summary — the user may read only that.\n\n```markdown\n# Project Profile — \u003Cname>\n\n\u003C2-3 sentence summary: what it is, what it does, how it is built.>\n\n**Stack** TypeScript 5.6 · React 19 · Vite 6 · Postgres 16 · Bun\n**Architecture** Feature-first, colocated tests, no barrel files\n**Entry points** `src\u002Fmain.tsx` (browser), `server\u002Findex.ts` (API)\n**Health** 1 conventions gap, 3 oversized files, no type-check in CI\n\n## 1. What This Is\n\n\u003Cpurpose, domain, who uses it, and what it is not>\n\n## 2. Stack and Runtime\n\n| Layer | Choice | Version | Evidence |\n| ----- | ------ | ------- | -------- |\n\n## 3. Structure\n\n\u003Cannotated tree of the meaningful directories, with a one-line purpose each>\n\n## 4. Module Boundaries\n\n\u003Cwhich units exist, their public surface, and the allowed dependency direction>\n\n## 5. Entry Points and Flow\n\n\u003Chow a request or command travels from entry to storage, naming the files>\n\n## 6. Conventions and Standards\n\n\u003Cthe conventions table from Phase 2, including what is enforced vs merely stated>\n\n## 7. Commands\n\n| Purpose | Command     | Source          |\n| ------- | ----------- | --------------- |\n| Dev     | `bun dev`   | package.json:12 |\n| Test    | `bun test`  | package.json:14 |\n| Lint    | `bun lint`  | package.json:15 |\n| Build   | `bun build` | package.json:16 |\n\n## 8. CI Gates\n\n\u003Cwhat CI enforces, and what it does not>\n\n## 9. Observations\n\n\u003Cthe non-obvious: architectural risks, oversized files, drift from documented\nrules, missing tests, dead code, inconsistencies worth a \u002Fs13n pass>\n\n## 10. Unknowns\n\n\u003Cwhat could not be determined, and the file or question that would resolve it>\n```\n\n## Output Rules\n\n- **Summarize; never dump.** Do not paste file contents or whole manifests. Extract and interpret.\n- **Every claim carries evidence.** A file path, a line number, or a command output.\n- **Separate declared from observed.** \"Documented: no inline comments\" vs \"Observed: 3 files use inline comments.\"\n- **Rank observations by what the user should do first.** A missing CI type-check outranks a stale comment.\n- **Keep it readable.** Tables over prose, a tree over a paragraph, and no wall of text. If a section has nothing, omit it rather than printing \"N\u002FA\".\n- **Flag genuine uncertainty as uncertainty.** Write \"could not determine\" and move on.\n- **Do not recommend a rewrite.** Describe the project as it is. Recommendations belong in a planning step the user asks for.\n\n## Persisting the Profile\n\nIf the user wants the model kept for later sessions, write it to `.claude\u002Fcontext\u002Fproject-profile.md` or a path they name, and tell them the path. Do not create a file unless asked — exploration must not leave a trace by default.\n\n## Depth\n\nScale the effort to the repository and the ask.\n\n| Repository                     | Approach                                                                                     |\n| ------------------------------ | -------------------------------------------------------------------------------------------- |\n| Small (under ~50 source files) | Read the manifests, the conventions docs, and most source files. Be exhaustive.              |\n| Medium (~50-500 files)         | Manifests, conventions, structure, one representative file per layer, and the largest files. |\n| Large (~500-5,000)             | Manifests, conventions, structure, and CI only. Ask the user to name a target for depth.     |\n| Monorepo                       | Workspace root plus the packages the user names, or one representative package.              |\n\n## Common Mistakes\n\n| Mistake                                                       | Why It Breaks                                                     | Correct Approach                                            |\n| ------------------------------------------------------------- | ----------------------------------------------------------------- | ----------------------------------------------------------- |\n| Skipping the conventions docs                                 | You describe a project and miss the rules that govern changing it | Read `AGENTS.md`, `CONTRIBUTING.md`, and lint configs first |\n| Reading files instead of structure                            | Burns enormous context for a fraction of the signal               | Read manifests, tree shape, and signatures                  |\n| Reporting a convention as enforced when it is only documented | The user plans on a gate that does not exist                      | Check whether CI or lint actually enforces it               |\n| Guessing the architecture from directory names                | Names lie; `services\u002F` is often not a service layer               | Confirm with import edges and one representative file       |\n| Claiming completeness after sampling                          | The user plans on a partial picture they believe is whole         | State what you skipped, explicitly                          |\n| Dumping file contents                                         | Unreadable and destroys the value of the summary                  | Extract, interpret, cite                                    |\n| Printing a `.env` file                                        | Leaks secrets into the transcript and the user's history          | List keys only; redact values                               |\n| Writing files unprompted                                      | Exploration should leave no trace                                 | Read-only unless asked                                      |\n| Recommending a rewrite                                        | Not what was asked, and it undermines trust in the findings       | Describe; let the user decide                               |\n| Reporting stale docs as fact                                  | Docs drift; the code is the truth                                 | Separate declared from observed, and note disagreement      |\n| One giant prose section                                       | Unreadable                                                        | Tables, trees, and a scannable structure                    |\n| Ignoring the commit history                                   | It reveals real priorities and convention drift                   | Read the recent log and the subject patterns                |\n",{"name":5,"description":7,"allowed-tools":24},"Read Grep Glob Bash","main","42203fde5eb692ba254e058168ee1af75855fcb4","skills\u002Fexplore",0,"https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills","skills.marvellofaisal.me","https:\u002F\u002Fgithub.com\u002FMarvellbrazil","marvellofaisal.me","Software Engineering Student | Backend & API Developer | Tech Enthusiast | Homelabber","2026-10-23 22:23:52",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":47},[37,40,43],{"tool":38,"command":39},"npx","npx skills add Marvellbrazil\u002Fmy-agentic-skills --skill explore",{"tool":41,"command":42},"bunx","bunx skills add Marvellbrazil\u002Fmy-agentic-skills --skill explore",{"tool":44,"command":45},"pnpm","pnpm dlx skills add Marvellbrazil\u002Fmy-agentic-skills --skill explore","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fexplore\u002FSKILL.md",true,[],[50,57,64,71,78,85],{"skillNo":51,"name":52,"nameZh":6,"description":53,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":19},"01M7DDQP2F4E42AMTGY0VCJ6G2","writing-guidelines","Review docs\u002Fprose for Writing Guidelines compliance. Use when asked to \"review my docs\", \"check writing style\", \"audit prose\", \"review docs voice and tone\", or \"check this page against the writing handbook\".","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-guidelines\u002FSKILL.md",[],[],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M7DDQM00XGSGT8V797C89R2E","writing-for-agents","Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwriting-for-agents\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7DDQEF1HP9CMH88HMN9MCBA","wrangler","Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fwrangler\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":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M7DDQA32ZAG7J4PMRXRSQ3NQ","workers-best-practices","Reviews and authors Cloudflare Workers code against production best practices. Load when writing new Workers, reviewing Worker code, configuring wrangler.jsonc, or checking for common Workers anti-patterns (streaming, floating promises, global state, secrets, bindings, observability). Biases towards retrieval from Cloudflare docs over pre-trained knowledge.","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fworkers-best-practices\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M7DDQ5482QSC0FC4NX107FVY","web-perf","Analyzes web performance using Chrome DevTools MCP. Measures Core Web Vitals (LCP, INP, CLS) and supplementary metrics (FCP, TBT, Speed Index), identifies render-blocking resources, network dependency chains, layout shifts, caching issues, and accessibility gaps. Use when asked to audit, profile, debug, or optimize page load performance, Lighthouse scores, or site speed. Biases towards retrieval from current documentation over pre-trained knowledge.","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fweb-perf\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7DDQ1GZ7Z2Y67460SJJEMC0","using-agent-skills","Discovers and invokes agent skills. Use when starting a session, or when you need to decide which skill or workflow applies to the piece of work at hand. This is the meta-skill that governs how all other skills are discovered and invoked.","https:\u002F\u002Fgithub.com\u002FMarvellbrazil\u002Fmy-agentic-skills\u002Fblob\u002Fmain\u002Fskills\u002Fusing-agent-skills\u002FSKILL.md",[],[]]