EVOKORE-MCP Skills Overview

EVOKORE-MCP indexes 266 SKILL.md files across 16 active categories, with the recursive SkillManager index seeing roughly 336 skill-related entries once supporting reference and asset files are included. This page gives the narrative shape of each category. Every category is tagged as Code workflow, Non-code workflow, or Both so a reader can match category to intent at a glance.

What this covers

• Category inventory with skill counts
• Per-category narratives with code / non-code workflow tags
• How to find a skill from the assistant side
• Operator notes on indexing, search ranking, and adapter skills

Category inventory

AWESOME CLAUDE CODE RESOURCES, MCP WRAPPERS, OFFICIAL MCP SERVERS, and upstream/ contain reference material and submodule vendoring but no top-level SKILL.md files; they are not counted above.

Category narratives

ARCHITECTURE

Code workflow. A single deep-refactor skill (improve-codebase-architecture) anchored on the bounded contexts in ADR-0005. Use it when surfacing architectural friction inside one context and turning shallow modules into deep ones.

AUTOMATION AND PRODUCTIVITY

Non-code workflow. Personal-productivity skills ported from the Anthropic skills cookbook: brand-guidelines, file-organizer, image-enhancer, invoice-organizer, slack-gif-creator, tailored-resume-generator, and the theme-factory family.

ANTHROPIC COOKBOOK

Non-code workflow. Custom skills from the Anthropic Cookbook: analyzing-financial-statements, applying-brand-guidelines, creating-financial-models, and the cookbook-audit notebook reviewer. Reference material (notebook walkthroughs, contribution guides) ships alongside but is not surfaced as SKILL.md.

COMMUNICATION

Both. The zoom-out adapter ported from mattpocock/skills. Use it when an agent needs to move up an abstraction layer and find the right seam for a change — equally applicable to a code refactor and to a non-code planning artifact.

CONTEXT

Code workflow. The ubiquitous-language adapter, hard-coupled to the eight bounded contexts in ADR-0005. Crystallizes a per-context GLOSSARY.md from replay/evidence JSONL plus src/ identifiers.

DEVELOPER TOOLS

Both. The largest non-wshobson category. Includes Anthropic's artifacts-builder, mcp-builder, skill-creator, webapp-testing; the document-skills suite (docx, pdf, pptx, xlsx); the reverse-engineering lane (ghidra-*, reverse-engineering-*, malware-triage-workflow, unknown-binary-onboarding, semantic-recovery-campaign, debugger-driven-analysis); plus frontend-design, setup-pre-commit, developer-growth-analysis, changelog-generator, and refly. The document-skills suite is non-code; everything else is code-side.

EVOKORE EXTENSIONS

Code workflow. Skills written for EVOKORE-specific runtime behavior: hooks-automation, sparc-methodology, verification-quality (truth scoring), agentic-jujutsu (concurrent claim-based writes), browser, github-release-management (canary rollouts), v3-mcp-optimization, anti-slop (tool-call review), and the master-workflow-555 driver.

GENERAL CODING WORKFLOWS

Code workflow. The day-to-day session workflows: arch-aep-runner, implementation-session, session-wrap, pr-manager, docs-architect, repo-ingestor, planning-with-files, security-review, and the phase-spec-optimizer retrospective tool.

HIVE FRAMEWORK

Code workflow. Goal-driven agent design (hive, hive-concepts, hive-create, hive-credentials, hive-debugger, hive-patterns, hive-test) plus the legacy triage-issue skill. Use the meta-router hive when starting a new agent project.

ORCHESTRATION FRAMEWORK

Both. The densest category. Top-level skills (handoff-protocol, policy-pack-v1, agent-archetypes, tool-governance, workflow-templates, schemas, coding-reference, improvement-cycles, aep-framework, panel-of-experts) plus the commands/ directory of orch-*, sparc-*, tdd, quality-gate, scope-lock, handoff, session-checkpoint, pattern-learn, context-budget, fleet-status, agent-spawn, verify-quality, workflow-run, and the analytics miners (session-retrospective-miner, cross-project-process-miner, narrative-quality-scorer). Panel of Experts and AEP cover both code and non-code artifacts; many orch-* commands are code-side.

PLANNING

Code workflow. Two evidence-driven planning skills: to-prd (synthesize a PRD from prior session evidence and panel-review it) and to-issues (decompose a PRD into AFK/HITL-tagged vertical-slice issues).

PROJECT MANAGEMENT

Code workflow. github-triage — a 7-state label-based GitHub issue state machine that reads issue bodies and evidence rather than waiting for maintainer triggers.

QA

Code workflow. triage-bug — triages bugs from session evidence (evidence-capture JSONL, replay JSONL, telemetry, repo-audit) and produces a TDD fix plan plus a docs/bugs/{slug}.md artifact.

RESEARCH AND CONTENT

Non-code workflow. Research and content workflows: competitive-ads-extractor, content-research-writer, lead-research-assistant, meeting-insights-analyzer, twitter-algorithm-optimizer.

Stitch Skills

Non-code workflow. The Stitch design pipeline: enhance-prompt, design-md, stitch-design, stitch-loop, shadcn-ui, react-components, remotion. Use stitch-loop for autonomous baton-passing website builds. The "non-code" tag reflects that the deliverable is a designed artifact even though the pipeline emits front-end code.

wshobson Plugins

Both. The deepest reference library. 146 SKILL.md files spanning backend (CQRS, sagas, microservices, event stores, Temporal), frontend and mobile (React state, Next.js App Router, Tailwind, iOS/Android design), cloud infrastructure (Terraform, Istio, Linkerd, multi-cloud, mTLS), CI/CD, Kubernetes (manifests, Helm, GitOps, security), security scanning (SAST, STRIDE, attack trees, threat mitigation), data engineering (Spark, dbt, Airflow), LLM application development (RAG, LangChain, embedding strategies, prompt patterns), Python development (a full Python style, testing, and packaging set), JavaScript and TypeScript, Rust and Go systems, blockchain and Web3, observability (Prometheus, Grafana, SLOs, distributed tracing), accessibility, payment processing, reverse engineering, conductor TDD workflows, agent-teams coordination, startup business analysis, framework migration, documentation generation, game development (Godot, Unity ECS), HR and legal compliance, shell scripting, and incident response. Mostly code-side; the startup business analysis, HR and legal compliance, and documentation-generation skills are non-code.

Finding a skill

Ask the assistant to invoke search_skills:

"Do you have a skill for PostgreSQL database design?"

For broader natural-language intents, use resolve_workflow:

"I need to wrap up this session and leave a clean handoff."

EVOKORE will semantically rank the best workflow matches, explain why they matched with Why matched: lines, and inject the top skills directly into the response.

Operator notes

• Skill indexing is recursive; the runtime no longer relies on the old shallow two-level walk.
• Search quality uses aliases, tags, frontmatter metadata, and semantic hints. resolve_workflow reranks to prefer actionable top-level skills over deep reference leaves.
• The composition graph (skill-graph.json) records nextSteps[] edges and is regenerated via npm run skill-graph.
• Adapter skills (vendored from upstream submodules) carry upstream, upstream-sha, upstream-path frontmatter. See SKILLS/DEVELOPER TOOLS/skill-creator/SKILL.md for the contract and SKILLS/COMMUNICATION/zoom-out/SKILL.md plus SKILLS/CONTEXT/ubiquitous-language/SKILL.md for canonical examples.

See also

• All Skills Crib Sheet — flat lookup table with one-line descriptions
• Panel of Experts — the multi-persona review framework many skills compose with
• Architecture: AEP System — the engineering cycle that drives the orchestration workflows
• Usage — day-to-day invocation patterns

Last verified: 2026-05-20