← 返回 Skills 市场
logikoma

Agent Life

作者 logikoma · GitHub ↗ · v1.2.0 · MIT-0
cross-platform ⚠ suspicious
261
总下载
1
收藏
0
当前安装
3
版本数
在 OpenClaw 中安装
/install agent-life
功能描述
Backup, sync, and restore agent memory and state to the cloud using the Agent Life Format (ALF). Use when asked to back up agent data, sync memory to the clo...
使用说明 (SKILL.md)

Agent Life — Backup, Sync, and Restore

The alf CLI backs up, syncs, and restores your agent's memory, identity, credentials, and workspace files to the cloud using the Agent Life Format (ALF). All commands output JSON to stdout. Progress goes to stderr. See https://agent-life.ai

Install

Download and install the alf binary from GitHub Releases:

# Option 1: Download, inspect, then run the install script (recommended)
curl -sSL https://raw.githubusercontent.com/agent-life/agent-life-adapters/main/scripts/install.sh -o install-alf.sh
cat install-alf.sh    # inspect the script
sh install-alf.sh     # run it

# Option 2: Direct binary download (no install script)
# See platform binaries at: https://github.com/agent-life/agent-life-adapters/releases/latest

Source code: https://github.com/agent-life/agent-life-adapters (MIT license, open source) Install script source: https://github.com/agent-life/agent-life-adapters/blob/main/scripts/install.sh

The install script detects your platform, downloads the binary, verifies the SHA256 checksum, and installs to /usr/local/bin/alf (or ~/.local/bin/alf without root). Stdout is JSON:

{"ok":true,"version":"v0.2.0","installed_version":"alf 0.2.0","path":"/usr/local/bin/alf","checksum_verified":true}

Verify: alf --version

Authenticate

Get an API key at https://agent-life.ai/settings/api-keys, then store it:

alf login --key \x3Cyour-api-key>

Output:

{"ok":true,"key_masked":"alf_sk_1...cdef","config_path":"/home/user/.alf/config.toml"}

Check Environment

Before syncing, run alf check to verify the environment and auto-discover the workspace:

alf check -r openclaw

The command auto-discovers the workspace path from ~/.openclaw/openclaw.json or ~/.alf/config.toml. To specify a workspace explicitly:

alf check -r openclaw -w /path/to/workspace

Key fields in the JSON output:

Field Type Meaning
ready_to_sync bool true if all prerequisites are met
workspace.path string Discovered or specified workspace path
workspace.source string How the path was found: flag, alf_config, openclaw.json, or default
issues array Problems found, each with severity, code, message, fix
alf.api_key_set bool Whether an API key is configured

If ready_to_sync is false, read issues[] for what to fix. Each issue has a fix field with the exact command or action to resolve it.

Core Workflows

Pre-flight check then sync (recommended)

check=$(alf check -r openclaw)
ready=$(echo "$check" | jq -r '.ready_to_sync')
ws=$(echo "$check" | jq -r '.workspace.path')

if [ "$ready" = "true" ]; then
    alf sync -r openclaw -w "$ws"
else
    echo "$check" | jq -r '.issues[] | "[\(.severity)] \(.message)\
  Fix: \(.fix)"' >&2
fi

First-time backup

Export creates a local .alf archive, then sync uploads it to the cloud:

alf export -r openclaw -w \x3Cworkspace>
alf sync -r openclaw -w \x3Cworkspace>

Export output:

{"ok":true,"output":"agent-export.alf","agent_name":"Atlas","alf_version":"1.0.0-rc.1","memory_records":47,"file_size":102400}

Sync output (first sync — full snapshot):

{"ok":true,"sequence":0,"delta":false,"changes":null,"snapshot_path":"/home/user/.alf/state/abc-snapshot.alf","no_changes":false}

Incremental sync

After the first sync, subsequent syncs upload only what changed:

alf sync -r openclaw -w \x3Cworkspace>

Output:

{"ok":true,"sequence":5,"delta":true,"changes":{"creates":2,"updates":1,"deletes":0},"snapshot_path":"/home/user/.alf/state/abc-snapshot.alf","no_changes":false}

Restore from cloud

Download the latest state and restore to a workspace:

alf restore -r openclaw -w \x3Cworkspace>

If multiple agents are tracked locally, specify which one:

alf restore -r openclaw -w \x3Cworkspace> -a \x3Cagent-id>

Output:

{"ok":true,"agent_id":"a1b2c3d4","agent_name":"Atlas","sequence":5,"runtime":"openclaw","memory_records":47,"workspace":"/home/user/.openclaw/workspace","warnings":[]}

Import an archive

Import an .alf file into a workspace without going through the cloud:

alf import -r openclaw -w \x3Cworkspace> backup.alf

Validate an archive

Check an .alf file against the ALF JSON schemas:

alf validate backup.alf

Output:

{"ok":true,"valid":true,"errors":[],"warnings":[]}

Common Errors and Fixes

Error / Issue Code Cause Fix
no_api_key No API key configured alf login --key \x3Ckey>
workspace_not_found Workspace directory doesn't exist Pass correct path: alf check -r openclaw -w /correct/path
no_memory_content No MEMORY.md and no memory/ directory Agent has no memories yet — nothing to sync
service_unreachable API endpoint not responding Check network; verify api_url in ~/.alf/config.toml
HTTP 401 Unauthorized Bad or revoked API key alf login --key \x3Cnew-key>
HTTP 409 Conflict Sequence mismatch during sync alf restore first, then sync again
HTTP 402 agent_limit Subscription agent limit reached Upgrade at https://agent-life.ai

Environment Status

Check full environment and service status:

alf help status

Output includes config_exists, api_key_set, service_reachable, tracked agents[] with last_synced_sequence, and agent_service_status[] with online and server_latest_sequence.

Full Reference

For complete flag documentation, JSON output schemas, and error codes:

Data and Privacy

This skill uploads agent data to the agent-life.ai cloud service. Here is exactly what is sent:

Uploaded: Memory records (daily logs, curated memory, project notes), identity files (SOUL.md, IDENTITY.md), principals (USER.md), workspace config files (AGENTS.md, TOOLS.md, etc.).

NOT uploaded: Actual secrets (API keys, tokens, passwords) are never read or transmitted. The alf CLI only exports credential metadata — service names and labels (e.g. "GitHub API key: configured") — never the secret values themselves. Session transcripts and chat history are not uploaded.

Review before uploading: You can inspect exactly what will be uploaded before any data leaves your machine:

alf export -r openclaw -w \x3Cworkspace>   # creates a local .alf archive
alf validate agent-export.alf           # check the archive structure

The .alf archive is a standard file you can inspect. Nothing is uploaded until you explicitly run alf sync.

Config files read: The alf CLI reads ~/.alf/config.toml (its own config) and ~/.openclaw/openclaw.json (to auto-discover the workspace path). These paths are declared in the skill's requires.config metadata. No other files outside the workspace directory are read.

Storage: All data is encrypted at rest (AES-256 via AWS KMS, per-tenant keys). Data is stored in AWS S3 (blobs) and Neon Postgres (metadata), both in the US.

Access: Only the authenticated user (API key holder) can read or delete their data. There is no shared access, no analytics on user data, and no third-party data sharing.

Deletion: Delete individual agents via the web dashboard at agent-life.ai or via DELETE /v1/agents/:id. Account deletion removes all data.

API key scope: The ALF_API_KEY authenticates to your agent-life.ai account. It can only access data belonging to that account. Keys can be revoked and rotated at https://agent-life.ai/settings/api-keys.

Privacy policy: https://agent-life.ai/privacy

Environment Variables

Variable Description
ALF_API_KEY API key for agent-life.ai (fallback if not in config)
ALF_HUMAN Set to 1 for human-readable output instead of JSON
ALF_INSTALL_DIR Override install directory for install.sh
ALF_VERSION Pin install.sh to a specific release

File Locations

Path Purpose
~/.alf/config.toml API key, API URL, default runtime and workspace
~/.alf/state/{agent_id}.toml Sync cursor (last sequence, timestamp)
~/.alf/state/{agent_id}-snapshot.alf Last snapshot for delta computation
安全使用建议
This skill appears to do what it says, but it will read and upload sensitive agent data (memory, identity, credentials, and workspace files) to the Agent Life cloud. Before installing or running it: 1) Verify you trust https://agent-life.ai and review their privacy/security policy for stored backups; 2) Inspect the install script and verify checksums against an authoritative source (don’t run a raw script without review); 3) Limit the ALF_API_KEY scope if possible and rotate the key after testing; 4) Audit what files/configs in your ~/.openclaw and workspace contain secrets and consider removing or encrypting secrets before backup; 5) If you need stricter guarantees, review the alf CLI source code (repo linked in SKILL.md) and test on a disposable agent/workspace first. If any of the above raises concerns (e.g., missing checksum verification, untrusted hosting, or the service’s retention/authorization rules are unacceptable), treat the skill as unacceptable for production use.
功能分析
Type: OpenClaw Skill Name: agent-life Version: 1.2.0 The 'agent-life' skill (SKILL.md) facilitates the exfiltration of agent state, including memory, identity, and workspace configuration, to an external cloud service (agent-life.ai). It promotes a high-risk installation method using 'curl | sh' to execute a remote script from GitHub and install a binary ('alf'). While the documentation explicitly claims to avoid uploading secrets and provides a 'Data and Privacy' section to build trust, the tool's core functionality is data exfiltration and it requires access to local configuration files (~/.openclaw/openclaw.json), representing a significant security risk.
能力评估
Purpose & Capability
The name/description (backup, sync, restore of agent memory/state) match the declared requirements: the alf CLI, ALF_API_KEY, and the OpenClaw config path (~/.openclaw/openclaw.json) are all reasonable and expected for a tool that discovers an agent workspace and uploads its state.
Instruction Scope
SKILL.md instructs running the alf CLI to check, export, sync, restore, import, and validate agent archives. Those commands will read agent workspace files and agent config (explicitly referenced) and upload agent identity, memory, credentials and workspace files to the Agent Life cloud — this is coherent with the stated purpose but involves transmission of highly sensitive data.
Install Mechanism
The skill is instruction-only (no bundled code). It points to an install script and GitHub Releases for the alf binary; this is a common pattern and the SKILL.md recommends inspecting the install script and verifies SHA256. Still, running install scripts from raw URLs carries risk unless you verify checksum/source and the checksum source is trustworthy.
Credentials
Required env var (ALF_API_KEY) and config paths (~/.alf/config.toml, ~/.openclaw/openclaw.json) are directly related to the backup/sync purpose. However, these grant access to sensitive data (agent identity and any credentials stored in the workspace/config). Requesting such access is proportionate to a backup tool but increases risk if the cloud service or API key is compromised.
Persistence & Privilege
The skill is not forced-always and uses normal model-invocation. It does not request persistent platform privileges or to modify other skills' configs; its requested presence is appropriate for a user-invoked backup/restore helper.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-life
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-life 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.2.0
Declare config paths in requires.config metadata. Remove curl|sh one-liner. Expand Data and Privacy section with credential handling details, pre-sync review workflow, and API key scope info.
v1.1.0
Fix frontmatter parsing (broken YAML, malformed metadata JSON). Add requires.env declaration for API key. Change install instructions to use GitHub Releases URL. Add Data and Privacy section. Add direct binary download option.
v1.0.0
Initial release: backup, sync, restore for OpenClaw agents
元数据
Slug agent-life
版本 1.2.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 3
常见问题

Agent Life 是什么?

Backup, sync, and restore agent memory and state to the cloud using the Agent Life Format (ALF). Use when asked to back up agent data, sync memory to the clo... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 261 次。

如何安装 Agent Life?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-life」即可一键安装,无需额外配置。

Agent Life 是免费的吗?

是的,Agent Life 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Agent Life 支持哪些平台?

Agent Life 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Agent Life?

由 logikoma(@logikoma)开发并维护,当前版本 v1.2.0。

💬 留言讨论