Agent SSD Guard — AI skill for Claude Code
Keep AI coding agent CLIs from grinding down your SSD — relocate regenerable write targets to a RAM disk.
How to install Agent SSD Guard
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open u-ichi/agent-ssd-guard and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Agent SSD Guard does
Keep AI coding agent CLIs from grinding down your SSD — relocate regenerable write targets to a RAM disk.
Alternatives in AI
- HolyClaude — AI coding workstation: Claude Code + web UI + 7 AI CLIs + headless browser + 50+ tools 2k ★
- Go Modern Guidelines — Help AI coding agents write modern Go 1.7k ★
- Botmux — Bridge Feishu/Lark to AI coding CLIs — Claude Code, Codex, Gemini, OpenCode… every DM, group or topic spawns i 1.2k ★
README
agent-ssd-guard
Keep AI coding agent CLIs from grinding down your SSD.
Relocates the write targets that can be regenerated if lost onto a RAM disk and symlinks the original paths to them. Targets are declared per agent and can be toggled individually. If there is not enough free memory, it does nothing and gets out of the way.
macOS only. Bash, no dependencies.
[日本語版 README](README.ja.md)
Why
Agent CLIs rewrite diagnostic logs, caches and runtime state on short intervals. Some rewrite **multi-megabyte files in full every few seconds even when the content has not changed by a single byte.** SSD write endurance is finite, and on Apple Silicon Macs the storage is soldered in.
Measured on Codex CLI 0.145.0, macOS, one session active, 421-second window:
| Target | Rate | What happens |
|---|---|---|
~/.codex/cache/remote_plugin_catalog/.json |
65.1 GB/day | A 7.7 MB JSON rewritten in full every 8.6s. 35 of 55 rewrites (64%) were byte-identical to what they replaced. |
~/.codex/models_cache.json |
3.6 GB/day | Rewritten in full every 6.7s |
~/.codex/cache/codex_apps_tools/.json |
2.0 GB/day | Rewritten in full every 60s |
~/.codex/logs_2.sqlite |
18.9 GB/day | Diagnostic log. The TRACE default is hardcoded and cannot be turned down by the user |
The plugin catalog wrote 302 MB over 7 minutes to carry two distinct contents totalling 15.3 MB. Reported upstream as [openai/codex#35401](https://github.com/openai/codex/issues/35401).
For contrast, the same method over 307 seconds with Claude Code and Grok active found **under 1.5 GB/day combined and zero byte-identical rewrites.** Their writes are appends to session transcripts — every byte written is new information.
**These numbers are environment-specific.** Plugin count, number of resident sessions and agent version move them by an order of magnitude. Measure your own before deciding — see [Measuring](#measuring).
Install
git clone https://github.com/u-ichi
Related Skills
Auto Guard
A cached auto command review guard for AI coding agents — rules, memory and LLM verdicts (allow/deny/ask) befo
Universal AI Config
Generate tool-specific AI config files from shared templates. Write your AI instructions, skills, and agents o
Tokenomy
Tokenomy — A surgical toolkit for AI coding CLIs. Cut tokens, keep context, build faster.
Claude Failover
Backup brain for Claude Code. When your Claude Max plan hits its limit or Claude is down, flip one command and
Claude Code Practices
Interactive AI installer for Claude Code: setup once, keep context across sessions, enforce critical rules wit
Agent Manager X
A macOS desktop app to monitor your Claude Code, Codex, OpenCode AI coding agents in real-time with voice or b
Related Agents
Orquestrador
Watches the machine's resources (RAM, CPU, disk, processes, ports). Invoke before starting a heavy build/dev s
Kb Extraction Worker
Turn ONE raw or vendored source file into one graphify {nodes, edges, hyperedges} extraction chunk and write i
9remote
Long-running AI coding sessions keep working after you leave the desk, but checking on them from a phone usual