agent-plugins

Bootstrap an LLM-maintained wiki for the current project. Use when a user asks to create, initialize, or refresh a project wiki, LLM wiki, QMD wiki, or codebase knowledge base for Claude Code, Codex, Pi, or OpenClaw.

Plan a feature, bugfix, refactor, or improvement with wiki research first. Searches project and main cross-project wiki context, synthesizes Past Knowledge, then delegates to Compound Engineering planning when available or writes a standalone plan outline.

Search QMD-indexed project and main cross-project LLM wikis for past patterns, decisions, solutions, and pitfalls before planning or implementation. Use before feature planning, bug fixing, refactoring, architecture work, or code changes when a project wiki may exist.

Report llm-wiki maintenance status. Use only when the user explicitly mentions llm-wiki and asks for its installed version, update availability, project wiki structure, consent state, or refresh queue health. Do not select this skill for generic plugin, documentation, project-context, or upgrade questions.

Upgrade an existing project's managed llm-wiki structure without rerunning broad wiki bootstrap. Use when a user asks to migrate, update, or upgrade project-local .llm-wiki scripts, hooks, or changelog structure after installing a newer llm-wiki release.

Bootstrap an LLM-maintained wiki for the current project. Use when a user asks to create, initialize, or refresh a project wiki, LLM wiki, QMD wiki, or codebase knowledge base for Claude Code, Codex, Pi, or OpenClaw.

Plan a feature, bugfix, refactor, or improvement with wiki research first. Searches project and main cross-project wiki context, synthesizes Past Knowledge, then delegates to Compound Engineering planning when available or writes a standalone plan outline.

Search QMD-indexed project and main cross-project LLM wikis for past patterns, decisions, solutions, and pitfalls before planning or implementation. Use before feature planning, bug fixing, refactoring, architecture work, or code changes when a project wiki may exist.

Report llm-wiki maintenance status. Use only when the user explicitly mentions llm-wiki and asks for its installed version, update availability, project wiki structure, consent state, or refresh queue health. Do not select this skill for generic plugin, documentation, project-context, or upgrade questions.

Upgrade an existing project's managed llm-wiki structure without rerunning broad wiki bootstrap. Use when a user asks to migrate, update, or upgrade project-local .llm-wiki scripts, hooks, or changelog structure after installing a newer llm-wiki release.

Bootstrap an LLM-maintained wiki for the current project. Use when a user asks to create, initialize, or refresh a project wiki, LLM wiki, QMD wiki, or codebase knowledge base for Claude Code, Codex, Pi, or OpenClaw.

Search QMD-indexed project and main cross-project LLM wikis for past patterns, decisions, solutions, and pitfalls before planning or implementation. Use before feature planning, bug fixing, refactoring, architecture work, or code changes when a project wiki may exist.

Upgrade an existing project's managed llm-wiki structure without rerunning broad wiki bootstrap. Use when a user asks to migrate, update, or upgrade project-local .llm-wiki scripts, hooks, or changelog structure after installing a newer llm-wiki release.

Plan a feature, bugfix, refactor, or improvement with wiki research first. Searches project and main cross-project wiki context, synthesizes Past Knowledge, then delegates to Compound Engineering planning when available or writes a standalone plan outline.

Report llm-wiki maintenance status. Use only when the user explicitly mentions llm-wiki and asks for its installed version, update availability, project wiki structure, consent state, or refresh queue health. Do not select this skill for generic plugin, documentation, project-context, or upgrade questions.

Complete SEO content workflow for Claude Code, Codex, Pi, and OpenClaw. Use only when the user explicitly asks for Agent SEO, agent-seo, a /seo:* command, or one of its named SEO modes: research, write, humanize, fact-check, optimize, analyze-existing, rewrite, scrub, data, or performance-review. Do not select it for generic writing or analytics discussion. The legacy scrub mode performs a read-only formatting audit. All four hosts can invoke Agent SEO by name; Claude Code also exposes the workflows as /seo:* commands.

Complete SEO content workflow for Claude Code, Codex, Pi, and OpenClaw. Use only when the user explicitly asks for Agent SEO, agent-seo, a /seo:* command, or one of its named SEO modes: research, write, humanize, fact-check, optimize, analyze-existing, rewrite, scrub, data, or performance-review. Do not select it for generic writing or analytics discussion. The legacy scrub mode performs a read-only formatting audit. All four hosts can invoke Agent SEO by name; Claude Code also exposes the workflows as /seo:* commands.

Complete SEO content workflow for Claude Code, Codex, Pi, and OpenClaw. Use only when the user explicitly asks for Agent SEO, agent-seo, a /seo:* command, or one of its named SEO modes: research, write, humanize, fact-check, optimize, analyze-existing, rewrite, scrub, data, or performance-review. Do not select it for generic writing or analytics discussion. The legacy scrub mode performs a read-only formatting audit. All four hosts can invoke Agent SEO by name; Claude Code also exposes the workflows as /seo:* commands.

A team of three writing voices working as rivals. Use when the user asks for Agent Writing, agent-writing, to investigate a topic across a project's data, to draft a story-first piece from a brief, to run a writer-editor cycle on a draft, or to run an end-to-end journalist → writer ↔ editor pipeline. The writer and editor are deliberately adversarial — cooperation creates sycophancy — and they run as a continuous cycle until the editor's verdict is `ready` or a configured maximum number of rounds is reached. Artifacts land in the user's project working directory under `./writing/`. In Codex, users invoke this skill by asking for Agent Writing; in Claude Code, the same workflows are also exposed as /write:* commands.

Clone your team's best code reviewers from their pull-request history and run them as a review panel. Use when the user wants to build or update a custom code reviewer from their team's review history, extract a reviewer's persona/style from past PR comments, mine GitHub review history for what reviewers care about, or review code in the voice of specific teammates. Ranks the prominent reviewers in a repo, collects each one's comments in context (the diff they were left on, the PR's intent), and extracts a persona per reviewer — what they care about, what they catch, what they let slide, how they sound — rather than a deterministic rule list. Each persona becomes a reviewer agent; run them together as a panel. In Codex this skill is the entry point; in Claude Code the same flows are also the /reviewer:extract, /reviewer:review, and /reviewer:update commands.

Clone your team's best code reviewers from their pull-request history and run them as a review panel. Use when the user wants to build or update a custom code reviewer from their team's review history, extract a reviewer's persona/style from past PR comments, mine GitHub review history for what reviewers care about, or review code in the voice of specific teammates. Ranks the prominent reviewers in a repo, collects each one's comments in context (the diff they were left on, the PR's intent), and extracts a persona per reviewer — what they care about, what they catch, what they let slide, how they sound — rather than a deterministic rule list. Each persona becomes a reviewer agent; run them together as a panel. In Codex this skill is the entry point; in Claude Code the same flows are also the /reviewer:extract, /reviewer:review, and /reviewer:update commands.

Clone your team's best code reviewers from their pull-request history and run them as a review panel. Use when the user wants to build or update a custom code reviewer from their team's review history, extract a reviewer's persona/style from past PR comments, mine GitHub review history for what reviewers care about, or review code in the voice of specific teammates. Ranks the prominent reviewers in a repo, collects each one's comments in context (the diff they were left on, the PR's intent), and extracts a persona per reviewer — what they care about, what they catch, what they let slide, how they sound — rather than a deterministic rule list. Each persona becomes a reviewer agent; run them together as a panel. In Codex this skill is the entry point; in Claude Code the same flows are also the /reviewer:extract, /reviewer:review, and /reviewer:update commands.

A team of three writing voices working as rivals. Use when the user asks for Agent Writing, agent-writing, to investigate a topic across a project's data, to draft a story-first piece from a brief, to run a writer-editor cycle on a draft, or to run an end-to-end journalist → writer ↔ editor pipeline. The writer and editor are deliberately adversarial — cooperation creates sycophancy — and they run as a continuous cycle until the editor's verdict is `ready` or a configured maximum number of rounds is reached. Artifacts land in the user's project working directory under `./writing/`. In Codex, users invoke this skill by asking for Agent Writing; in Claude Code, the same workflows are also exposed as /write:* commands.

A team of three writing voices working as rivals. Use when the user asks for Agent Writing, agent-writing, to investigate a topic across a project's data, to draft a story-first piece from a brief, to run a writer-editor cycle on a draft, or to run an end-to-end journalist → writer ↔ editor pipeline. The writer and editor are deliberately adversarial — cooperation creates sycophancy — and they run as a continuous cycle until the editor's verdict is `ready` or a configured maximum number of rounds is reached. Artifacts land in the user's project working directory under `./writing/`. In Codex, users invoke this skill by asking for Agent Writing; in Claude Code, the same workflows are also exposed as /write:* commands.

Retrieve Screenote pages, screenshots, annotations, and private crops, then comment after applying a fix.

Capture an explicit HTTP(S) page at desktop, tablet, or mobile viewports and publish private local files through the Screenote JSON CLI.

Discover approved HTTP(S) application routes and publish serial per-route captures through allowlisted screenshot create calls.

Retrieve Screenote pages, screenshots, annotations, and private crops, then comment after applying a fix.

Capture an explicit HTTP(S) page at desktop, tablet, or mobile viewports and publish private local files through the Screenote JSON CLI.

Discover approved HTTP(S) application routes and publish serial per-route captures through allowlisted screenshot create calls.

Retrieve Screenote pages, screenshots, annotations, and private crops, then comment after applying a fix.

Capture an explicit HTTP(S) page at desktop, tablet, or mobile viewports and publish private local files through the Screenote JSON CLI.

Discover approved HTTP(S) application routes and publish serial per-route captures through allowlisted screenshot create calls.