Suggest Qoder to update checkup feature to clean unused skills/MCP to reduce credits consume

Feature Requests
1

Problem

Qoder currently has no equivalent of Claude Code’s doctor / checkup command. Over time, projects accumulate skills and MCP servers that are never used, haven’t been used in a long time, or are used extremely
rarely. These entries are continuously injected into the system prompt (skill name + description lists, MCP tool schemas), which:

  1. Wastes input tokens — every request pays for describing unused capabilities, directly increasing billing cost.
  2. Degrades LLM attention — a long list of irrelevant skills/MCP tools distracts the model from the capabilities actually relevant to the current project, hurting tool-selection accuracy and response
    quality.
  3. Is invisible to the user — there is no built-in way to see “what’s loaded, what’s actually used, and what’s dead weight,” so users have no basis for deciding what to disable or remove.

Solution

Add a /doctor (or /checkup) command that audits installed skills and MCP servers against real usage data and offers guided cleanup:

  1. Inventory & usage report
    • List all skills and MCP servers configured for the current project (and user-level ones affecting it).
    • For each entry, show: name, one-line summary of its function, and usage statistics (times invoked, last-used date, invocation frequency) tracked locally from session history.
    • Classify entries: actively used / rarely used / long unused (e.g., >30/90 days) / never used.
  2. Recommendations
    • Flag entries that are candidates for disabling or removal (never used, or no usage within a configurable window), with an estimated token/context cost each one contributes.
  3. Agent-assisted cleanup
    • After the user reviews and approves (per-entry or batch), the agent performs the disable/remove operations — editing the appropriate settings files for skills and MCP server configs.
    • Finally, report: what was disabled/removed, and the current remaining skills/MCP list with the resulting context/token footprint reduction.

Use Case

  • A project has been in use for months; dozens of skills and MCP servers have been added incrementally. The user runs /doctor to see that 15 of 30 skills were never invoked and 2 MCP servers haven’t been used
    in 90 days, approves cleanup, and cuts unnecessary prompt overhead.
  • Before a long coding session on a cost-sensitive budget, the user wants to verify that only relevant capabilities are loaded.
  • After onboarding a new plugin pack or MCP bundle, the user wants to confirm which pieces are actually being used a few weeks later.

Priority

  • :red_circle: High - Blocking issue
  • :yellow_circle: Medium - Important improvement
  • :green_circle: Low - Nice to have

Additional Info

  • Prior art: Claude Code’s doctor / checkup command, which inspects environment health; this request extends the concept to capability-hygiene auditing.
  • Usage tracking can be implemented locally (e.g., recorded in session/telemetry storage already available to the CLI); no server-side changes strictly required.
  • Suggested related UX: a summary line in session status showing the current total context cost of loaded skills/MCP, so the problem stays visible between /doctor runs.