Skip to content
Back to skills

Codelens

ASecurity

Use the existing CodeLens MCP server for multi-file architecture analysis, change review, call-path tracing, impact checks, and language diagnostics. Use when Codex needs indexed structural evidence across files or modules. Do not use for a single exact text lookup or an already-local one-file edit.

  • 3 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added October 11, 2026
ai-agents

Works with

  • mcp

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned October 11, 2026

npx -y skills add mupozg823/codelens-mcp-plugin --skill codelens --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Codelens?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Codelens
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/mupozg823-codelens/badge)](https://www.skillsdirectory.com/skills/mupozg823-codelens)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: codelens
description: Use the existing CodeLens MCP server for multi-file architecture analysis, change review, call-path tracing, impact checks, and language diagnostics. Use when Codex needs indexed structural evidence across files or modules. Do not use for a single exact text lookup or an already-local one-file edit.
---

# CodeLens

Use CodeLens inside Codex's native loop. The plugin supplies this workflow only; the
host-level `codelens` MCP registration remains the source of truth.

## Workflow

1. Resolve the absolute repository root. If CodeLens tools are deferred, discover
   `codelens prepare_harness_session` with native tool search.
2. Make the first CodeLens call
   `prepare_harness_session(project=<absolute-root>, detail="compact")`.
   Confirm the effective project and reuse that binding for the session.
   Declare observed host capabilities and available tool names when known;
   model names alone do not establish host capabilities.
3. Select the smallest facade that answers the task:

   - `review`: architecture, changed-file, boundary, dead-code, or duplicate analysis.
   - `graph`: callers, callees, type hierarchy, request trace, or changed-file references.
   - `diagnose`: file, symbol, unresolved-reference, or issue diagnostics.
   - `get_capabilities`: index, LSP, semantic, and runtime readiness checks.
   - `search` / `overview`, when exposed: bounded symbol/context retrieval and file structure.

4. Call only tools exposed to this host; use native tool search for deferred tools.
   If a needed tool is unavailable, use native evidence instead of repeating discovery.
   Keep results bounded and reuse existing evidence; cache hits still add host rounds.
   Use native `rg` for exact text and native edit tools for mutations.
5. Cite file and line evidence, state index or diagnostic limitations, and run the
   narrowest relevant verification after a change.

If the `codelens` MCP dependency or a diagnostic capability is unavailable, continue
with native tools and report the limitation. Do not start a second server or invent
a fallback endpoint. Bootstrap again only after a binding or connection change.

Files in this skill

  • SKILL.md2.1 KB
  • agents/openai.yaml345 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.