agent-browser
Browser automation CLI for AI agents. Use when the user needs to interact with websites,…
Audit architecture, dead code, duplication, type confusion, and code smells across a codebase. Excludes formatting, performance profiling, security audits, and feature-level reviews.
$ npx -y skills add pedronauck/skills --skill architectural-analysis --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/architectural-analysisContext preview
The summary Claude sees to decide when to auto-load this skill.
Audit architecture, dead code, duplication, type confusion, and code smells across a codebase. Excludes formatting, performance profiling, security audits, and feature-level reviews.
name: architectural-analysis description: "Audit architecture, dead code, duplication, type confusion, and code smells across a codebase. Excludes formatting, performance profiling, security audits, and feature-level reviews." disable-model-invocation: true metadata: author: Pedro Nauck github: https://github.com/pedronauck repository: https://github.com/pedronauck/skills
Audit the requested scope and report concrete design problems. An audit alone is read-only; existing explicit authorization to apply findings can include remediation. Use the project's boundaries and public compatibility policy when judging a design.
1. Resolve scope from the request/context. Map entry points and relevant dependencies with `rg --files`, scoped `rg` searches, and available static-analysis output. Do not create one todo per source file or assume a TypeScript stack. 2. Prioritize signals: dead exports, duplicated responsibilities, cycles, confused ownership/types, and expensive change patterns. Read the matching section of `references/detection-catalog.md` when classifying a candidate; thresholds are clues, not automatic defects. 3. Trace usage before declaring code dead, including dynamic loading, reflection, framework hooks, tests, and published APIs. An unreferenced public export is not safe to delete solely because this checkout has no consumer. 4. Confirm each material finding in source. Explain the behavior or maintenance cost, cite paths/lines, state uncertainty, and propose the smallest useful change. Avoid a separate generic smell sweep when it adds no evidence. 5. Match report size to the request. For a broad audit, use `assets/report-template.md` as an adaptable outline and save `.audits/architectural-analysis-<timestamp>.md`; distinguish inspected and uninspected areas. For a narrow answer, a concise finding list is enough. Missing coverage must never be reported as `None found`. 6. Summarize the important findings and evidence. Existing source and summary templates are optional presentation aids, not two required reports.
A curated collection of 87 agent skills for Claude Code and compatible AI coding assistants — 40 original (⭐️), 34 hand-picked (💎), and 13 marketing & business (📣).
Repo: pedronauck/skills
Browser automation CLI for AI agents. Use when the user needs to interact with websites,…
Answer questions about the AI SDK and help build AI-powered features. Use when developers:…
Autonomously optimize any Claude Code skill by running it repeatedly, scoring outputs against…
Centrifugo real-time messaging server expert for WebSocket PUB/SUB, channel management, JWT…
Retrieves up-to-date documentation, API references, and code examples for any developer…
Diátaxis Documentation Expert. An expert technical writer specializing in creating…