About claude-code-ops-router
claude-code-ops-router is a routing skill for Claude Code operations. It maps a requested operation to a specific specialist skill and requires resolving that skill from the active router’s canonical location.
Key features
- 01Routes operations to specialist skills
- 02Maps requests to explicit skill paths
- 03Resolves paths from the active router
- 04Requires reading selected instructions in full
- 05Separates routing from unrelated Claude workflows
Use cases
- 01Troubleshooting a missing Claude Code skill
- 02Changing a Claude Code statusline
- 03Analyzing Claude usage and quota
- 04Maintaining a plugin marketplace
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ daymade/ claude-code-skills --skill claude-code-ops-routerAfter installing, type/claude-code-ops-routerin Claude Code to confirm it loaded.
Details
From the project READMEClaude Code operations router
Choose the one specialist that owns the requested operation. Locate this
loaded claude-code-ops-router/SKILL.md in the active catalog or invocation,
follow symlinks to its canonical path, and verify its parent is
claude-code-ops-router. Its parent's parent is the suite root. Resolve the
selected relative path against that directory and verify it is a direct
sibling SKILL.md inside the root. Never anchor resolution to the task's
working directory or a remembered plugin cache. ${CLAUDE_PLUGIN_ROOT} may
help in Claude Code, but it is absent in Codex. Stop if the active router path
cannot be identified unambiguously.
Read the selected child SKILL.md in full before acting. Continue a
truncated read and load its task-required references. Its instructions,
scripts, and assets remain installed; only its automatic catalog entry is
hidden. Claude users retain the original manual commands named by each child's
frontmatter. The export-repair directory claude-export-txt-better exposes
/daymade-claude-code:fixing-claude-export-conversations.
| Requested operation | Read this exact file |
|---|---|
| Repair a Claude Code plugin or Skill that is installed but missing, disabled, or failing to activate | ../claude-skills-troubleshooting/SKILL.md |
| Install, change, or repair the Claude Code statusline | ../statusline-generator/SKILL.md |
Repair line wrapping, tables, paths, or tool output in an exported Claude Code .txt conversation | ../claude-export-txt-better/SKILL.md |
| Create or maintain a Claude Code plugin marketplace, suite membership, or marketplace manifest | ../marketplace-dev/SKILL.md |
| Analyze Claude Code/Desktop Code token use, cost, cache, quota burn, or 5-hour blocks | ../claude-usage-analyst/SKILL.md |
Configure isolated Claude Code model-provider profiles, aliases, source-backed Claude/Codex Skill activation, or context-window size; reduce preloaded MCP schema context after active-profile and provider checks, or safely suppress selected MCP servers; repair early compaction, [1m] display, or CLAUDE_CODE_MAX_CONTEXT_TOKENS | ../claude-switch-models-setup/SKILL.md |
| Migrate existing Claude personal memory into tool-agnostic AGENTS.md reference documents, or retire a project's auto memory so documents own everything it held | ../claude-migrate-memory-to-doc/SKILL.md |
| Set a one-shot local ping after a Claude subscription quota reset to start the next usage window | ../claude-code-ping-start-5h-quota/SKILL.md |
Route by the requested operation, not by the word “Claude.” For example,
usage diagnosis selects claude-usage-analyst; a requested future reset ping
selects claude-code-ping-start-5h-quota. A Skill count, discovery-policy
audit, or cold-catalog change belongs to skill-governance when installed;
plugin activation repair is the narrower troubleshooting row here.
This router does not replace the direct entries for conversation-history reading and continuation, prior-work retrieval, hooks, CLAUDE.md editing, Lark, Claude.ai web-conversation export, terminal color screenshots, relayed web-search repair, or technology selection. Do not use it to handle those requests merely because Claude Code is mentioned. Selecting a child does not authorize an external installation, credential operation, or scheduled job; follow the selected child's instructions and the user's authorization.