← Back to Skills Marketplace
Agent Attestation
by
Ivan Cetta
· GitHub ↗
· v3.0.2
458
Downloads
0
Stars
1
Active Installs
13
Versions
Install in OpenClaw
/install agent-attestation
Description
Portable reputation system for agents v3 - Ed25519 signatures, input validation, handoff KV
Usage Guidance
This package appears to implement what it claims (local attestation, Ed25519 signing, and a handoff KV) and contains no network calls or requests for unrelated credentials, but it stores private keys and identity/reputation data unencrypted in default relative directories. Before installing or running: 1) review and run the code in a sandbox/isolated environment; 2) set keys_dir and KV directory to a secure absolute path outside any shared workspace; 3) restrict filesystem permissions (e.g., chmod 600 on private key files); 4) consider adding encryption at rest or protecting private keys with a passphrase or OS keyring; 5) install a vetted version of the 'cryptography' library from PyPI and review dependency versions; 6) audit the code yourself (or ask the author) if you need assurance there are no network endpoints beyond the included files. If you cannot ensure secure storage for the generated keys, treat the skill as risky for production use.
Capability Analysis
Type: OpenClaw Skill
Name: agent-attestation
Version: 3.0.2
The skill bundle implements an agent attestation and identity persistence system. While the code uses strong cryptography (Ed25519) in `attestation_system_v3.py` and includes explicit security warnings in `attestation_system_v3.py`, `example_usage.py`, and `handoff_kv.py` about storing unencrypted private keys and identity data on disk, this practice constitutes a significant vulnerability. The direct storage of unencrypted private keys and sensitive identity information on the filesystem, even with warnings, represents a high-risk capability that could lead to compromise if the designated storage directories (`./keys`, `./attestation_kv`) are not adequately secured by the operating environment. This falls under the 'suspicious' category as a vulnerability, not malicious intent, due to the inherent risk of the capability itself.
Capability Assessment
Purpose & Capability
Name/description (attestation, Ed25519, handoff KV) align with the included code: key management, signing/verification, input validation, and a simple on-disk KV are implemented. Minor inconsistency: registry metadata described the skill as instruction-only yet the package contains multiple Python source files implementing the features; this is likely just an author packaging choice rather than malicious misdirection.
Instruction Scope
SKILL.md and the example usage clearly require Python and the 'cryptography' library and warn about filesystem access. Runtime code only performs local file I/O (keys, attestations, manifest) and canonical signing/verification and input checks. There are no network calls or references to unrelated system paths or environment variables in the provided files.
Install Mechanism
No install spec or remote downloads are present; this is a code-bundle you run locally. The only external dependency is the Python 'cryptography' library (imported), which must be installed in your environment; there are no opaque third-party download URLs or extract steps.
Credentials
The skill requests no environment variables or external credentials, which is proportionate to its purpose. However, it generates and writes Ed25519 private keys to disk in PEM form unencrypted (KeyManager.generate_keypair) and stores identity/reputation data in an on-disk KV in cleartext by default. Default relative paths (./keys, ./attestation_kv) increase risk in shared or cloud/workspace environments. This is a storage/operational security concern (not an evidence of exfiltration).
Persistence & Privilege
The skill does not request forced persistent inclusion (always: false) nor does it modify other skills. It writes persistent files to configurable directories and thus can provide long-lived identity/reputation state; ensure those directories are chosen and permissioned appropriately. Autonomous invocation (disable-model-invocation: false) is the platform default and not a unique red flag here.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-attestation - After installation, invoke the skill by name or use
/agent-attestation - Provide required inputs per the skill's parameter spec and get structured output
Version History
v3.0.2
- Minor update with internal code changes.
- Modified: attestation_system_v3.py.
- Modified: handoff_kv.py.
- No changes to public API or documentation.
v3.0.1
- Updated documentation and removed change log notes from SKILL.md.
- Minor edits to supporting files for clarity and consistency.
- No changes to overall feature set or functionality.
v3.0.0
No file changes detected for version 3.0.0.
- No updates or modifications included in this release.
- All features, usage, and documentation remain unchanged from the previous version.
v2.0.6
- Updated SKILL.md structure for improved metadata and formatting.
- Explicitly added version, author, and keywords fields.
- Metadata now includes an emoji for openclaw compatibility.
- Content and user instructions remain unchanged.
v2.0.5
- Updated SKILL.md to clarify version notation (changed "v3.0" to "v3" in the "What's New" section and file descriptions).
- Generalized references to v3 throughout documentation; removed redundant ".0" in version listings.
- No code or functionality changes—documentation only.
v2.0.4
- Updated SKILL.md version reference in the description from "v3.0" to "v3"
- Set the version in metadata to string format ("3.0.0")
- No functional or code changes; documentation formatting and metadata consistency improvements only
v2.0.3
- SKILL.md documentation was updated for greater conciseness and clarity.
- Introductory and descriptive text at the top of the document was removed.
- All feature and usage information is unchanged; no code or logic updates.
v2.0.2
Major update: Agent-attestation v3.0 introduces cryptographic signatures, input validation, and persistent identity features.
- Added Ed25519 cryptographic signature support for attestations (attestation_system_v3.py).
- Introduced input validation for higher data integrity.
- Added HandoffKV for key/value storage to persist and restore identities (handoff_kv.py).
- Expanded SKILL.md with updated usage, key management, and input validation instructions.
- Existing v2.0 implementation and file compatibility maintained.
v2.0.1
Fix: filepath Path conversion, compute_web_of_trust dict bug
v2.0.0
v2.0: Rolling window (30 days), vouching with skin in the game, task weights (low/medium/high/critical), hybrid ERC-8004 ready
v1.1.1
Re-publish to trigger VirusTotal scan
v1.1.0
v1.1.0: Added example_usage.py, clarified ALPHA status, fixed crypto/email limitations in docs
v1.0.0
Agent Attestation 1.1.0 introduces a portable reputation system for agents, enabling the creation, verification, and sharing of attestations that persist beyond individual platforms.
- Create, verify, and share agent attestations with plain text reasons for vouching.
- Export and import attestations as portable JSON files (.attest extension).
- Compute basic web-of-trust reputation scores using direct and indirect attestations.
- Provides full Python implementation and example usage scripts.
- Designed for interoperability and resilience in decentralized agent networks.
Metadata
Frequently Asked Questions
What is Agent Attestation?
Portable reputation system for agents v3 - Ed25519 signatures, input validation, handoff KV. It is an AI Agent Skill for Claude Code / OpenClaw, with 458 downloads so far.
How do I install Agent Attestation?
Run "/install agent-attestation" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Attestation free?
Yes, Agent Attestation is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Agent Attestation support?
Agent Attestation is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Attestation?
It is built and maintained by Ivan Cetta (@nantes); the current version is v3.0.2.
More Skills