How does zlog differentiate between 'stale' logs and my actual conversation history?

zlog targets high-volume telemetry files like `.trace`, `.out`, and `.log` generated by AI coding tools, while explicitly whitelisting and protecting critical files like SQLite databases and `.jsonl` conversation histories.

Does this skill work with specific AI agents like Claude Code or Cursor?

Yes, zlog includes specialized path detection for Claude Code, Cursor, Aider, Windsurf, LM Studio, and Ollama, ensuring it finds logs even when they are hidden in deep application data folders.

What happens if a compression task fails mid-process?

zlog uses a transactional 'verify-before-delete' workflow: it compresses the file, verifies the archive's integrity, and ensures the file wasn't modified during the process before removing the original.

Which operating systems are supported by the zlog scripts?

zlog is cross-platform; it includes optimized POSIX shell scripts for macOS, Linux, and WSL, as well as native PowerShell scripts designed for Windows 11 environments.

What is included in the skill package upon purchase?

You receive the full set of production-ready scripts for all supported platforms, a configuration guide for custom pathing, and a library of pre-defined agent log locations.

Do I need to install any heavy third-party dependencies to run zlog?

While zlog is designed to be lightweight and portable, it achieves the best results if your system has `zstd`, `xz`, or `gzip` installed, though it will fallback to standard compression if necessary.

Works with the AI Agents you already use

+22 more

+18 agents

zlog

Compress stale AI agent logs and traces while protecting conversation history and databases.

Free

Security scanned

zlog

Example session with this skill installed

Run a preview to see how many stale logs from Cursor and Claude Code can be compressed.

Read your context and instructions
Compiled the zlog

I have scanned the known AI agent log directories for Cursor and Claude Code. I found several stale .trace and .log files that are candidates for compression. No files were modified during this preview.

Log Compression Preview

Agent	File Path	Size	Age	Status
Cursor	`~/Library/Logs/Cursor/2023-10-12_trace.log`	42.5 MB	3 days	Candidate
Claude	`~/.claude/logs/session_882.out`	12.1 MB	1 day	Candidate
Claude	`~/.claude/logs/session_881.out`	8.4 MB	2 days	Candidate
Cursor	`~/Library/Logs/Cursor/current.log`	1.2 MB	10s	SKIPPED (Active)
Aider	`~/.aider/logs/analytics.jsonl`	0.5 MB	5 days	PROTECTED

Summary of Candidates:

Total Files: 3
Uncompressed Size: 63.0 MB
Estimated Savings: ~85% (based on zstd-15)

Next steps

Run clean to compress these 3 candidate files and reclaim space.
Run clean --older-than 7 if you only want to compress logs from last week.
Run deep-preview to search for logs in recursive subdirectories.

Connects securely to your tools. The creator never sees your data.

About this skill

AI agent logs from tools like Claude Code, Cursor, and Aider quickly bloat your storage with multi-megabyte .trace and .out files. Manually hunting for these in hidden directories is tedious, and broad cleanup scripts risk deleting your conversation history or breaking live databases.

What it does

Log compression packs stale .log, .out, and .trace files using zstd, xz, or gzip to reclaim space.
Safety filtering protects SQLite databases, JSONL transcripts, and config files from being touched.
Integrity validation verifies every archive before deleting the source file to prevent data loss.
Lock detection automatically skips files currently being written by active agent sessions.
Dry-run previews lists exactly which files are candidates for compression without making changes.

How it works

Identify environment by detecting the OS and shell to select the correct POSIX or PowerShell engine.
Scan roots using a library of known AI agent paths to find stale session data.
Execute preview to show a list of candidates, their ages, and potential targets for compression.
Apply cleanup to compress files, verifying that the resulting archive is smaller than the source.

Frameworks & tools

This skill uses POSIX shell scripts (zstd, xz, gzip, tar) for Linux, macOS, and WSL, and native PowerShell scripts for Windows 11. It targets logs from Claude Code, Cursor, Gemini, Codex, Ollama, Windsurf, LM Studio, and Aider.

Why this beats prompting it yourself

Writing a shell one-liner to find and zip files often results in broken symlinks or deleted conversation history. This skill implements a strict transactional logic: it tests archive integrity and checks if the source file changed during compression before finalizing any deletion.

Use cases

Post-session cleanup to pack logs after a long day of coding with AI agents.
Storage maintenance to find and compress logs older than 7 or 30 days.
Disk recovery when hidden .trace files from background agents consume GBs of space.

Known limitations

Does not discover unknown agents in non-standard paths unless deep scan is used. Cannot compress SQLite databases or active chat transcripts. Requires zstd, xz, or gzip installed for maximum compression efficiency.

How to use

Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

Reviews

No reviews yet

Be one of the first to try it. Every listed skill passes our trust checks below.

Security scanned

Passed our 8-point scan before listing

Fresh listing

Recently published to Agensi

Free forever

No account required to browse

Trust & safety

Security scanned

Verified clean 4 days ago

Free to download with an account

Requires

GitHub

Agentskills

Shields

Skills

macOS

Node.js

Windows

Git

Listed4 days ago

zlog