erfana

Refresh project documentation after recent work – syncs docs/, README, CHANGELOG, AGENTS.md, CLAUDE.md, API specs and decision records with the current state of the code. Use after finishing a task or as housekeeping.

Translate a GitHub issue into a plain-language brief for a Product Owner / Project Manager / Business Analyst.

Researched multi-lens code review – fans out lens reviewers over a target, each grounded in recent cited best practices, and synthesizes one plain-language, severity-ranked report (PM/PO-friendly, with full technical detail kept for engineers).

Generate a stakeholder-facing project status summary in executive-brief format following the Pyramid Principle. The reader is a Product Owner / Project Manager / Business Analyst, not a developer. Translate technical artifacts (commits, branches, file diffs, hashes) into outcome language (milestones, capabilities, deliverables, queue items).

Generate a stakeholder-facing summary of the current Claude Code session in executive-brief format following the Pyramid Principle. The reader is a Product Owner / Project Manager / Business Analyst, not a developer. Translate technical artifacts (commits, branches, file diffs, hashes) into outcome language (milestones, capabilities, deliverables, queue items). Source content primarily from this conversation's context with a light read-only git probe for grounding.

Architecture reviewer for SOLID principles and design quality. MUST BE USED for technical architecture reviews. Use PROACTIVELY when evaluating code quality, structure, or design decisions. <example> Context: New module or feature was just implemented user: "Review the new payment module for architecture quality" assistant: "I'll use the architecture-reviewer agent to evaluate SOLID compliance, design patterns, and anti-patterns in the payment module." <commentary>Post-implementation architecture review requires systematic SOLID analysis – trigger architecture-reviewer.</commentary> </example> <example> Context: User wants to understand codebase health user: "How much technical debt do we have in the services layer?" assistant: "I'll use the architecture-reviewer agent to inventory technical debt markers and assess architectural health of the services layer." <commentary>Technical debt assessment is a core architecture-reviewer capability – trigger proactively.</commentary> </example> <example> Context: User suspects inconsistent patterns user: "Are we using design patterns consistently across the codebase?" assistant: "I'll use the architecture-reviewer agent to audit pattern usage, consistency, and appropriateness across the codebase." <commentary>Design pattern consistency audit requires systematic review – trigger architecture-reviewer.</commentary> </example>

MUST BE USED to draft a full article when the managing-articles orchestrator hands over an approved brief, outline, and research results and needs a coherent draft written to disk. Delegate here to draft every section to its outline word target, assemble the sections into one flowing draft, and format citations consistently - all in a single authoring pass. Use whenever the brief, outline, research-results, and sources files exist and the article now needs to be written, not when evidence is still being gathered. <example> Context: The orchestrator has an approved brief, an outline, and a sourced research-results file, and now needs the article written. user: "Draft the managed-Postgres comparison article from the approved outline." assistant: "The brief, outline, research-results, and sources are ready. I'll delegate to the article-drafter agent to write each section, assemble the draft, and format citations in one pass." <commentary>Section drafting, assembly, and citation formatting are one cohesive authoring task; the orchestrator hands this agent the four file paths, the language array, the output directory, and the version number.</commentary> </example> <example> Context: A bilingual article is ready to draft and the orchestrator needs independent Polish and English drafts, not a machine translation. user: "Napisz dwujezyczny artykul o RODO na podstawie zatwierdzonego konspektu." assistant: "The brief specifies Polish and English. I'll delegate to the article-drafter agent to write independent per-language drafts that each carry their citations." <commentary>The agent honors the language array and writes draft-v{version}.pl.md and draft-v{version}.en.md as independent drafts, so the orchestrator routes bilingual authoring here.</commentary> </example>

MUST BE USED to design a hierarchical article outline when the managing-articles orchestrator has both an article brief and a research-results file and needs a structured H2/H3 plan with per-section word-count targets before drafting. Delegate here to group research themes into a logical section hierarchy, assign a word-count target to each section that sums to the brief's total, map supporting sources to every section, and emit a single shared outline file. Use whenever evidence has been gathered but the article structure has not yet been planned. <example> Context: The orchestrator holds an article brief and a completed research-results file and now needs a structure before any drafting begins. user: "We have the brief and the research for the managed-Postgres comparison. Plan the article." assistant: "Brief and research-results are both ready. I'll delegate to the article-outliner agent to build an H2/H3 outline with per-section word targets and source mapping." <commentary>Structuring gathered evidence into a section hierarchy with word budgets is exactly this agent's job; the orchestrator hands it the brief path, the research-results path, and an output directory.</commentary> </example> <example> Context: A bilingual Polish-and-English article is planned and the orchestrator needs one shared outline before per-language drafts. user: "Zaplanuj strukture dwujezycznego artykulu o RODO - mamy juz brief i research." assistant: "The brief specifies Polish and English. I'll delegate to the article-outliner agent to produce one shared outline.md for both languages with section word targets." <commentary>The outline is shared across languages, so the orchestrator routes the single structuring pass here and the bilingual drafts reuse the one outline.</commentary> </example>

MUST BE USED to conduct and organize web research for an article when the managing-articles orchestrator needs sourced, themed findings written to a results file. Delegate here to run searches and fetches under trust controls, structure findings by theme with a citation per claim, de-duplicate, and emit a single research-results file plus a source list. Use whenever an article brief and research questions exist but the underlying evidence has not yet been gathered, organized, and analyzed. <example> Context: The orchestrator has produced an article brief and a list of open research questions and now needs evidence gathered before drafting. user: "Write an article comparing managed Postgres providers for European startups." assistant: "I have the brief and research questions ready. I'll delegate to the article-researcher agent to gather, organize, and analyze sourced findings into a research-results file." <commentary>Evidence-gathering with citations and trust controls is exactly this agent's job; the orchestrator hands it the brief path, the questions, and an output directory.</commentary> </example> <example> Context: A bilingual article is planned and the orchestrator needs both Polish and English sources collected and kept separate. user: "Przygotuj dwujezyczny artykul o RODO dla zespolow produktowych." assistant: "The brief specifies Polish and English. I'll delegate to the article-researcher agent to research both languages and return per-language source lists." <commentary>The agent honors the language array, researches Polish-language sources, and keeps per-language source lists, so the orchestrator routes bilingual evidence work here.</commentary> </example>

MUST BE USED to run a single editorial review pass over article drafts when the managing-articles orchestrator needs categorized findings before publication. Delegate here to review structure, style, fact-verification, and publication requirements in one pass, applying language-conditional metrics and the critical fact rule, and to return severity-ranked findings the orchestrator compiles into a review report. Use whenever drafts, a brief, research results, and a source list exist and the article has not yet been editorially reviewed. <example> Context: The orchestrator has a finished draft plus the brief, research results, and sources, and needs an editorial review before it can compile a report. user: "Review the managed-Postgres article draft before we publish." assistant: "Drafts and supporting files are ready. I'll delegate to the article-reviewer agent to check structure, style, facts, and publication requirements and return categorized findings." <commentary>One read-only pass over structure, style, facts, and compliance is exactly this agent's job; the orchestrator hands it the file paths and compiles the returned findings into the report.</commentary> </example> <example> Context: A bilingual article has Polish and English drafts and the orchestrator needs each reviewed against its own language metrics. user: "Sprawdz dwujezyczny artykul o RODO przed publikacja." assistant: "Both drafts and the sources file are present. I'll delegate to the article-reviewer agent to review each language against its own metrics and return per-language findings." <commentary>The agent honors the language array and applies Polish signals to the Polish draft and English targets to the English draft, so the orchestrator routes bilingual review here.</commentary> </example>

Use when the brief is vague, the visual style is unset, or the user asks which direction to pick.

Use when the user wants an infographic, data visualization, or info-dense chart.

Use when the user wants an animation, motion graphic, animated explainer, or video export.

Use when the user wants a clickable UI prototype, app mockup, or interactive design exploration.

Use when the user explicitly asks for a critique, review, score, or feedback on completed design work.

4 hooks across 3 events