Skip to content
Back to skills

Update Overview

ASecurity

Inventory project Codex instructions, skills, agents, rules, and configuration key names, and update a local overview without accessing runtime or account state.

  • 32 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 1, 2026
ai-agentsshelldatabase

Security analysis

A100/100

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

Scanned October 1, 2026

npx -y skills add kochetkov-ma/claude-brewcode --skill update-overview --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Update Overview?

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

Security grade badge for Update Overview
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/kochetkov-ma-update-overview/badge)](https://www.skillsdirectory.com/skills/kochetkov-ma-update-overview)

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: update-overview
description: Inventory project Codex instructions, skills, agents, rules, and configuration key names, and update a local overview without accessing runtime or account state.
---

# Project Codex overview

Parse the RU/EN prompt for mode and target. A single literal `status`, `show`, or `update` wins; otherwise show/list/check and покажи/список/проверь resolve to read-only `status`, update/refresh and обнови/освежи resolve to `update`. Empty input defaults to `update`. Extract paths from prose; clarify outcome-changing ambiguity. Emit one English plan with mode, project scope, output path, and result before work.

Inventory only the current project's `AGENTS.md`, `.codex/rules/`, `.codex/skills/`, `.codex/agents/`, `.agents/skills/`, and explicitly named project Codex config files. Discover via focused `rg --files`; record path, format, purpose, symlink target, and missing references. Follow links only to tracked repository instruction/reference sources; record external targets without reading them. Rules require explicit instruction references; discovery aliases do not imply plugin activation.

For JSON/TOML configuration, report structural key names only, never values. Do not dump config contents or parser diagnostics containing source lines. Do not read `.env`, credentials, auth, sessions, histories, databases, caches, shell snapshots, secret environment values, or global runtime directories. Do not scan or measure excluded directories.

`status` returns findings without writes. `update` writes the requested project-local document, or `.codex/reports/<YYYYMMDD-HHMMSS_codex-overview>/overview.md` if none was named. Preserve unrelated content when updating an existing overview. Include instruction/rule relationships, skills and agents, config key inventory, source paths, inspection date, and gaps; use a compact tree only for allowed paths. Summarize changed inventory and written paths. Never synchronize to global profiles or `.claude/`.

Use the current tracked inventory as evidence; distinguish missing references from unavailable external sources. Do not import Claude runtime scans, config dumps, plugin commands, or global synchronization into this workflow.

Files in this skill

  • LICENSE47 B
  • SKILL.md2.2 KB

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.