OpenClaw skills that give AI agents permanent memory and on-chain identity on the Autonomys Network.
autonomys/
├── auto-memory/ # Permanent storage and memory chains (primary)
├── auto-drive/ # ⚠️ Deprecated — renamed to auto-memory
└── auto-respawn/ # On-chain identity and memory anchoring
Each skill lives in its own directory under autonomys/ with a SKILL.md that defines the skill interface, installation steps, and usage instructions.
clawhub.ai/0xautonomys/permanent-memory
npx clawhub install permanent-memoryPermanent decentralized storage via the Autonomys Auto Drive API. Gives your agent:
- Permanent file storage — upload any file to the Autonomys Network and get back a CID. The data is distributed across the network and never expires.
- Memory chains — save identity, knowledge, and key decisions as a linked list where each entry points to the previous one via CID. A future instance can reconstruct who it was from a single CID.
- Local CID tracking — the skill maintains a local state file (
automemory-state.json) with the latest head CID, chain length, and timestamp. This means auto-memory is fully functional on its own — no wallet, no tokens, no on-chain transactions needed.
auto-memory only needs an Auto Drive API key (free at ai3.storage). It's the right starting point for any agent that wants permanent memory.
clawhub.ai/0xautonomys/respawn
npx clawhub install respawnOn-chain identity and memory anchoring on the Autonomys Network. Gives your agent:
- Permanent identity — a wallet with two addresses (consensus
su...and EVM0x...) derived from a single recovery phrase. This is the agent's verifiable on-chain identity. - Token management — check balances, transfer tokens, and bridge between the consensus layer and Auto-EVM.
- On-chain CID anchoring — write the latest memory chain head CID to the MemoryChain smart contract. This makes the head CID recoverable from any machine using just the agent's EVM address — no local state needed.
auto-respawn requires auto-memory (or equivalent) to have something worth anchoring. It completes the resurrection loop: auto-memory builds the chain, auto-respawn anchors the pointer on-chain, and a new agent instance can recover everything from just its address. The reverse dependency does not apply — auto-memory works independently with local CID tracking.
The original name for the auto-memory skill. Preserved for backward compatibility but no longer maintained. Use autonomys/auto-memory/ instead.
GitHub Actions runs on PRs and pushes to main (.github/workflows/ci.yml):
- TypeScript — type check (
tsc --noEmit), lint (eslint .), and test (vitest run) for auto-respawn (Node 22) - Shell —
shellcheck -x -S warningon all auto-memory scripts (except_lib.sh, which is validated in context via-x)
Skills are published to ClawHub via .github/workflows/publish.yml. The workflow triggers on tags matching */v*.
-
Tag the commit using the pattern
<skill>/v<semver>:git tag auto-memory/v0.2.0 git tag auto-respawn/v0.1.0-beta.3
-
Push the tag:
git push origin auto-memory/v0.2.0
-
The workflow automatically:
- Parses the skill name and version from the tag (strips the
vprefix — ClawHub expects bare semver like0.2.0, notv0.2.0) - Resolves the skill directory (
autonomys/auto-memoryorautonomys/auto-respawn) - Validates that
SKILL.mdexists in the directory - Publishes to both ClawHub accounts (primary
0xautonomysand legacyjim-counter)
- Parses the skill name and version from the tag (strips the
| Tag prefix | Directory | Primary slug (0xautonomys) | Legacy slug (jim-counter) | Display name |
|---|---|---|---|---|
auto-memory/v* |
autonomys/auto-memory |
permanent-memory |
auto-memory |
Auto Memory |
auto-respawn/v* |
autonomys/auto-respawn |
respawn |
auto-respawn |
Auto Respawn |
CLAWHUB_TOKEN_0XAUTONOMYSrepository secret — for the primary0xautonomysaccountCLAWHUB_TOKENrepository secret — for the legacyjim-counteraccount- The tagged commit must contain a valid
SKILL.mdin the skill's directory