MC

mcp-integration

Creates Storybook stories. Use this to generate varied component states and edge cases.

Install

mkdir -p .claude/skills/mcp-integration-siamese001 && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/18849" && unzip -o skill.zip -d .claude/skills/mcp-integration-siamese001 && rm skill.zip

Installs to .claude/skills/mcp-integration-siamese001

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Use this skill when choosing among repository-configured MCP servers, diagnosing MCP availability, or coordinating a cross-server workflow. Use adg-sqlite directly for structural dependency analysis.
199 chars✓ has a “when” trigger
Advanced

Key capabilities

  • Classify user intent and select a primary authority
  • Confirm server configuration and tool exposure
  • Inspect current tool schema for parameters
  • Prefer read-only health or discovery calls before state-changing calls
  • Apply least privilege, bounded results, explicit timeouts, and rollback for mutations

How it works

This skill routes user requests to configured MCP servers based on intent, ensuring proper tool exposure and applying security best practices for interactions.

Inputs & outputs

You give it
User intent requiring interaction with an MCP server
You get back
Interaction with the selected MCP server, reporting the server/tool used and any fallback

When to use mcp-integration

  • Create new component stories
  • Generate component edge case tests
  • Define component args and controls
  • Setup Storybook accessibility

About this skill

MCP integration

Root .mcp.json is the configured-server source of truth. A declared server is not proven callable until its tools are exposed and a suitable read-only health or discovery call succeeds in the current session.

Routing table

NeedProcedure
Filesystem batches or cross-root readssections/01-filesystem-mcp.md
Redis projection and cache statesections/02-redis-cache.md
Third-party GitHub repository Q&Asections/03-deepwiki.md
Versioned external package documentationsections/04-context7.md
Browser interaction and UI evidencesections/05-playwright.md
Semantic similarity retrievalsections/06-vector-db.md
Notion page or database operationssections/07-notion.md
Web search, extraction, mapping, or crawlingsections/08-tavily-research.md
Runtime spans, anomalies, or telemetrysections/09-otel-telemetry.md
Structured pytest discovery and runssections/10-pytest-mcp.md
Git, branch, PR, or issue operationssections/11-gitkraken.md
Optional cross-session knowledge graphsections/12-memory-mcp.md
Explicit durable task trackingsections/13-task-manager-mcp.md

Load only the section needed for the current intent. Do not recreate per-server redirect skills in the active skill tree.

Workflow

  1. Classify the user intent and select one primary authority.
  2. Confirm the server is configured and the expected tool is currently exposed.
  3. Inspect the current tool schema; do not infer parameters from a stale skill example.
  4. Prefer a read-only health or discovery call before a state-changing call.
  5. Apply least privilege, bounded results, explicit timeouts, and a rollback or compensating action for mutations.
  6. Report the actual server/tool used and any degraded fallback.

Boundaries

  • Structural imports, consumers, layers, and blast radius belong to adg-sqlite/graph-analysis.
  • Runtime traces do not replace static dependency evidence, and static graphs do not prove runtime execution.
  • Vector similarity does not establish a code dependency.
  • Native file tools are preferred for ordinary single-file workspace reads and writes.
  • File memory under memory/ remains project memory authority; a memory MCP is an optional graph projection.
  • A dormant section is documentation for restoration, not evidence that a server is callable.

Read agents-tier1-companion.md for the compact operator view and SUPPORTING.md only for legacy overflow that has not yet been retired.

Validation

python .codex/governance/scripts/sync_mcp_config.py --check
python ops_scripts/ci/check_agents_mcp_coverage.py
python ops_scripts/ci/run_skill_contract_gates.py

When not to use it

  • For structural imports, consumers, layers, and blast radius analysis
  • When runtime traces are expected to replace static dependency evidence
  • When vector similarity is expected to establish a code dependency

Limitations

  • Does not handle structural imports, consumers, layers, or blast radius analysis
  • Runtime traces do not replace static dependency evidence
  • Vector similarity does not establish a code dependency

How it compares

This skill provides a structured approach to interacting with various MCP servers, ensuring proper configuration, tool schema inspection, and secure practices, unlike direct, unmanaged server calls.

Compared to similar skills

mcp-integration side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
mcp-integration (this skill)015dReviewAdvanced
writing-registry-meta56moNo flagsBeginner
tsx01moReviewBeginner
React + Ant Design Frontend04moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry