name: agent-life description: 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 cloud, restore from cloud, or migrate agent state. version: 1.8.0 metadata: openclaw: requires: env: - ALF_API_KEY bins: - alf config: - ~/.alf/config.toml - ~/.openclaw/openclaw.json primaryEnv: ALF_API_KEY envVars: - name: ALF_API_KEY required: true description: API key for agent-life.ai (fallback if not in config). - name: ALF_HUMAN required: false description: Set to 1 for human-readable stdout instead of JSON. - name: ALF_HOME required: false description: Override the home base for ~/.alf and ~/.openclaw (config, sync state, vault). Falls back to $HOME when unset. Set this if the agent's $HOME is unstable. - name: ALF_INSTALL_DIR required: false description: Override install directory used by install.sh. - name: ALF_VERSION required: false description: Pin install.sh to a specific release. - name: ALF_ALLOW_UNVERIFIED required: false description: Set to 1 to allow install.sh to proceed when the SHA256 checksum cannot be verified. Default is to fail. homepage: https://agent-life.ai

Agent Life — Backup, Sync, and Restore

The alf CLI backs up, syncs, and restores a precise allowlist of your agent's memory, identity, and configuration files to the agent-life.ai cloud using the open Agent Life Format (ALF). Credential secrets are encrypted client-side with an offline vault key before they leave your machine — agent-life.ai sees only ciphertext. You can preview exactly what will be uploaded with --dry-run and further restrict the set with a workspace-local .alfignore file. 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 from GitHub Releases, requires a successful SHA256 checksum verification, and installs to /usr/local/bin/alf (or ~/.local/bin/alf without root). Stdout is JSON:

{"ok":true,"version":"v0.1.8","installed_version":"alf 0.1.8","path":"/usr/local/bin/alf","checksum_verified":true}

Checksum verification is mandatory by default. A hash mismatch always aborts the install (exit 4) and cannot be overridden. If verification cannot be performed at all — the .sha256 file is missing or empty, or no sha256sum/shasum tool is available — the script also exits 4 by default; set ALF_ALLOW_UNVERIFIED=1 to opt out of that case only (not recommended), in which case checksum_verified is false and a warnings array in the output JSON records the reason.

For production use, pin to a specific release:

ALF_VERSION=v0.1.8 sh install-alf.sh

Verify: alf --version

Authenticate

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

alf login --key <your-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:

If ready_to_sync is false, read issues[] for what to address. Each issue has a suggestion field with human-readable guidance — display it to the user; do not pipe it to a shell.

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)\n  Suggestion: \(.suggestion)"' >&2
fi

Preview what will be uploaded (--dry-run)

Before the first backup or any sensitive change, list the exact files that would be included without writing an archive or contacting the cloud:

alf export -r openclaw -w <workspace> --dry-run

Output:

{"ok":true,"dry_run":true,"agent_name":"Atlas","memory_records":47,"files":[{"path":"SOUL.md","size":2048},{"path":"IDENTITY.md","size":1024},{"path":"memory/2026-01-15.md","size":4096}],"excluded_by_alfignore":3,"total_size":102400}

Use this to confirm the inclusion set matches your expectations before any data leaves the machine.

First-time backup

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

alf export -r openclaw -w <workspace>
alf sync -r openclaw -w <workspace>

Export output:

{"ok":true,"output":"agent-export.alf","agent_name":"Atlas","alf_version":"1.0.0-rc.1","memory_records":47,"file_size":102400,"excluded_by_alfignore":0}

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 <workspace>

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 safely

Restoring downloads cloud state into a workspace directory. To avoid overwriting a live workspace, use one of the safe-restore paths below before the destructive command.

Option 1: Dry run — download and validate the archive without touching the workspace:

alf restore -r openclaw -w <workspace> --dry-run

Output lists the files that would be written, without creating them:

{"ok":true,"dry_run":true,"agent_id":"a1b2c3d4","sequence":5,"would_write":[{"path":"SOUL.md","size":2048},...],"warnings":[]}

Option 2: Point-in-time preview — fetch a specific historical sequence and inspect it without updating local sync state. ~/.alf/state is not touched, so a later regular alf sync is unaffected:

alf restore -r openclaw -w <workspace> --at-sequence 5

Option 3: Restore to a fresh path first — restore to a scratch directory, inspect it, then promote it:

alf restore -r openclaw -w /tmp/restore-preview
diff -r /tmp/restore-preview <real-workspace>   # inspect differences
# if satisfied: cp -r /tmp/restore-preview/* <real-workspace>/

Restore from cloud (destructive)

Once you have previewed the restore, run the destructive command. It writes into -w <workspace> and creates the directory if it does not exist:

alf restore -r openclaw -w <workspace>

If multiple agents are tracked locally, specify which one:

alf restore -r openclaw -w <workspace> -a <agent-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 <workspace> 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":[]}

Controlling What's Uploaded — .alfignore

By default, alf export includes a fixed allowlist of workspace files (see Data and Privacy below). To narrow the set further, drop a .alfignore file at the workspace root. The syntax is the same as .gitignore:

# Exclude an entire subdirectory of memory/
memory/private/

# Exclude a single file
HEARTBEAT.md

# Re-include one file from an excluded directory
!memory/private/keep-this.md

Rules:

• Patterns are relative to the workspace root.
• Lines starting with # are comments; blank lines are ignored.
• Negation (!pattern) re-includes a previously-excluded path.
• .alfignore itself is never uploaded.
• The agent's vault file at ~/.alf/vault/credentials.json is outside the workspace and is not affected by .alfignore. Use alf vault to control vault contents.

Inspect the effect with alf export --dry-run — the excluded_by_alfignore count and the files[] list reflect the filtered set.

Common Errors and Fixes

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:

• Agent-readable: https://agent-life.ai/docs/cli.md
• Human-readable: https://agent-life.ai/docs/cli

Data and Privacy

This skill uploads agent data to the agent-life.ai cloud service. The set is precise and inspectable.

Exactly what is uploaded

From the workspace root, an exact allowlist of 8 files (each only if present):

• SOUL.md
• IDENTITY.md
• AGENTS.md
• USER.md
• TOOLS.md
• HEARTBEAT.md
• BOOTSTRAP.md
• MEMORY.md

Plus the workspace's memory/ directory, recursively. Symlinks inside memory/ are not followed, so files outside the workspace cannot be pulled in through a link.

Plus the agent's ALF vault at ~/.alf/vault/credentials.json (when present). Vault records are encrypted client-side; only ciphertext leaves the machine.

A .alfignore file at the workspace root, if present, removes paths from this set. Run alf export --dry-run to confirm the final list.

Exactly what is NOT uploaded

• Your vault key — never leaves your machine, never transmitted, never derivable from anything we send.
• Plaintext credential secrets — encrypted client-side before any network call.
• Files outside the allowlist (any workspace file other than the 8 named root files or contents of memory/).
• Files reachable only via symlinks pointing outside the workspace.
• Runtime keystores (e.g. OpenClaw auth-profiles.json).
• Session transcripts and chat history.
• The .alfignore file itself.

Credential encryption (end-to-end)

Credential secrets are encrypted on your machine using XChaCha20-Poly1305 (default) or AES-256-GCM with a vault key you control. The vault key is generated and stored offline — agent-life.ai never receives it and cannot decrypt your credentials. Non-sensitive metadata (service name, label, capability tags, timestamps) is stored alongside the ciphertext for UX and auditing. Lose the vault key and the encrypted credentials are unrecoverable; back it up the same way you would back up an SSH private key. See https://agent-life.ai/cli#alf-vault for vault key generation and recovery.

Install integrity

The alf binary is downloaded from GitHub Releases and SHA256-verified during install. A hash mismatch always aborts the install (exit 4) and is never bypassable. If verification cannot complete — missing .sha256 file, empty checksum, or no sha256sum/shasum available — the script also exits 4 by default; the user must explicitly opt out with ALF_ALLOW_UNVERIFIED=1 to install without verification, in which case the JSON output reports "checksum_verified":false and a warnings array. Inspect the install script before running it (see Install above).

Review before uploading

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

alf export -r openclaw -w <workspace> --dry-run   # list files without writing
alf export -r openclaw -w <workspace>             # write local .alf archive
alf validate agent-export.alf                     # check the archive structure

Nothing is uploaded until you explicitly run alf sync.

Files read on your machine

The alf CLI reads the following local files. No other files on the filesystem are read.

Under your home directory:

• ~/.alf/config.toml — the CLI's own config (API key, API URL, defaults).
• ~/.alf/state/{agent_id}.toml — local sync cursor. Read on every sync, never uploaded.
• ~/.alf/state/{agent_id}-snapshot.alf — last snapshot, used to compute deltas. Read on every sync, never uploaded.
• ~/.alf/vault/credentials.json — the encrypted credential vault. Read during export; only ciphertext leaves the machine (see Credential encryption above).
• ~/.openclaw/openclaw.json — read to auto-discover the workspace path for OpenClaw runtimes.

Inside the workspace:

• The 8 root files in the upload allowlist plus memory/ recursively (see Exactly what is uploaded above).
• .alfignore at the workspace root, if present — read to filter the export set; never uploaded.

The requires.config metadata in this skill's frontmatter declares only the two preexisting config files the skill expects (~/.alf/config.toml and ~/.openclaw/openclaw.json); the state and vault paths are managed by alf itself and are created on first use.

Storage

All data is encrypted at rest on agent-life.ai (AES-256 via AWS KMS, per-tenant keys), in addition to the client-side encryption already applied to credential payloads. 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.

Retention and deletion

Data is retained until you delete it. There is no automatic expiry. Delete individual agents via the web dashboard at agent-life.ai or via DELETE /v1/agents/:id. Account deletion removes all data associated with the account. Local data (the .alf archive, ~/.alf/state/, and the vault key) is your responsibility — back up the vault key and any local archives the same way you would back up SSH keys.

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/agents/api-keys.

Privacy policy

https://agent-life.ai/privacy

Environment Variables

File Locations

这个 Skill 质量较好，文档结构清晰、说明详细，安全考虑也很周全（加密、预览、忽略规则）。主要优点是工作流覆盖完整，从安装到恢复都有指引。不足之处是缺少实际使用示例，对新手不够友好，且依赖外部 CLI 工具安装，可能存在兼容性问题。整体适合有一定技术基础的用户使用。