akuroglo
UserThe quality layer for Claude Code. Not everything you need — what survived seven months of daily use: self-enforcing guards (claim-guard, scope-lock, bash guard) + field-proven workflow skills, selection telemetry published (zeros included). Plugin wires 6 hooks; the rest is opt-in. 44 self-tests.
Categories
Indexed Skills (10)
doc-feature
Generate feature/module documentation after implementation. Covers business context (why, for whom, use cases), technical architecture (data flow, API, decisions), and operational details (limitations, edge cases). Integrates with the SSOT structure (docs/DOCUMENTATION_STRATEGY.md). USE WHEN: after >3 new files or a feature change. Triggers: "document this feature", "update docs/features", "doc-feature", "sync docs ↔ code", «задокументируй фичу», «обнови docs/features».
product-discovery
Product discovery and prioritization framework. Use when planning features, validating assumptions (OST), prioritizing work (RICE), analyzing jobs-to-be-done (JTBD), or testing hypotheses before building. Triggers: "prioritize", "what to build", "RICE", "JTBD", "OST", "validate assumption", "opportunity sizing", "feature priority", "should we build", "discovery", «приоритизируй», «что строить»
systematic-debugging
Four-phase root cause debugging methodology. Use when investigating bugs, errors, or unexpected behavior. Prevents guess-and-check thrashing. Triggers: bug, error, debug, investigate, «баг», «ошибка», «не работает», «почему падает», «сломалось».
config-audit
Audit Claude Code configuration against current best practices. Use when checking configuration health, after updates, or periodically to ensure optimal setup. Triggers on: 'audit config', 'check setup', 'config health'.
deep-research
Multi-source research with mechanical quality control. USE WHEN a question needs outside information and the answer has to be trustworthy: technology evaluation, market or competitive analysis, architecture decisions, "what do practitioners actually do", or any topic with 3+ dimensions. Handles narrow questions (sequential, one context) and broad ones (parallel independent agents) through the same quality gate. Триггеры: «deep-research», «дипресёрч», «глубокое исследование», «сравни технологии/подходы», «как это делают на практике», «research».
deploy-setup
Install and adapt this verification bundle (claim-guard, scope-lock, bash guard) into the current project. Use when the user says 'install this setup', 'set up claude code for this project', 'add these hooks', 'deploy the guardrails', or points at the claude-code-setup repo and asks to apply it. Reads the project first, asks a few questions, shows a plan, and only then writes.
doc-routing
Documentation Router & Architect — decides WHERE new information belongs in the project documentation system. Use BEFORE creating/updating any docs: PRD, features, journeys, integrations, NFR, requirements. Guarantees consistency and prevents duplication. Triggers: "doc routing", "document this", "where to put docs", "doc architecture", «куда писать», «документация», "/doc-routing"
pr-hero
Presentation-engineering pipeline — turns topic + core idea into a complete, defensible deck spec: audience strategy, narrative arc, evidence-tiered arguments, slide-by-slide design brief, anticipated Q&A, and an adversarial QA gate. Facts are researched and cited (exa / tavily / WebSearch), never invented. Emits a renderer-agnostic DeckSpec plus human-readable briefs. Use when the user asks to make a presentation / deck / pitch, «сделай презентацию про», «собери питч», "pr-hero", "/pr-hero".
rpi-methodology
RPIV (Research-Plan-Implement-Verify) workflow with chat command triggers. Use when planning features, organizing work, or following a structured development methodology. Триггеры: «новая фича», «спланируй/план», «proceed/ok», «research [тему]», new feature, plan, R→P→I→V cycles.
verify-before-claim
Gate against asserting unverified facts. USE WHEN about to claim that something "doesn't exist" / "is blocked" / "isn't deployed", state a measured value/number, describe how an API behaves, or assert a vendored component's capability — especially from memory or several turns ago. Triggers: "doesn't exist", "is blocked", "already done", "the metric equals", any claim about API/enum/payload, "verify", ��не существует», «заблокировано», «уже сделано», «перепроверь».
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.