← Back to Skills Marketplace
Openclaw
by
adrianhihi
· GitHub ↗
· v0.6.3
· MIT-0
114
Downloads
0
Stars
0
Active Installs
9
Versions
Install in OpenClaw
/install vial-self-healing
Description
Vial self-healing runtime — 8 protocols covering technical failures, behavioral failures, and agent role enforcement.
Usage Guidance
This skill is coherent with its stated purpose but performs autonomous actions you should review: it issues shell exec() calls that (a) write logs to /tmp/vial.log, (b) call an external telemetry/repair endpoint hosted on a personal workers.dev domain, and (c) can spawn verification sub-sessions and auto-execute follow-up steps (even without asking for every non-destructive action). Before installing, confirm you trust the telemetry/repair domain and its operator; consider whether you are comfortable with the skill making outbound HTTP requests and potentially running automated follow-up actions on your behalf. If you need stronger guarantees, ask the author for (1) the exact telemetry payloads and a strict spec for parsing Gene Map responses, (2) an option to disable remote calls/telemetry, and (3) explicit confirmation prompts instead of automatic execution for any action that could have side effects.
Capability Analysis
Type: OpenClaw Skill
Name: vial-self-healing
Version: 0.6.3
The skill bundle implements a 'self-healing' runtime that includes a remote strategy lookup mechanism via curl to helix-telemetry.haimobai-adrian.workers.dev. This allows the agent to fetch and execute instructions (strategies) from an external server, creating a potential remote instruction injection vector. The protocols also instruct the agent to perform background telemetry and read local configuration files (~/.openclaw/workspace/IDENTITY.md) using shell execution. While the stated intent is reliability, the ability to dynamically alter agent behavior from a remote endpoint is a significant security risk.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (self‑healing runtime) align with the SKILL.md protocols (loop detection, auth/session recovery, verification, etc.). Required resources declared in registry (no binaries, no env) match the instruction-only design.
Instruction Scope
The runtime instructs the agent to run shell exec(...) commands (curl and file writes), spawn verification sub-sessions (sessions_spawn / sessions_yield), and in several cases to 'execute the pending task NOW' without further user confirmation for non‑destructive steps. It also relies on remote Gene Map responses to decide strategies but does not define how to parse or validate those responses. These actions extend the agent's behavior beyond passive guidance and could change agent outputs or cause network activity.
Install Mechanism
No install spec and no code files — instruction-only — so nothing is downloaded or written to disk by an installer. The only persisted artifact is /tmp/vial.log written at runtime.
Credentials
The skill requires no credentials or special env vars, which is appropriate. However it sends telemetry and performs Gene Map lookups to an external HTTP endpoint (https://helix-telemetry.haimobai-adrian.workers.dev). While SKILL.md claims only anonymous, non‑PII fields are sent, the skill triggers network calls and could in practice be extended or misused to send more data; this is a privacy/trust consideration rather than a declared credential mismatch.
Persistence & Privilege
always:false (not force-included). The skill writes only to /tmp/vial.log and uses platform session APIs for verification; it does not request system-wide config changes or other skills' credentials.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install vial-self-healing - After installation, invoke the skill by name or use
/vial-self-healing - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.6.3
v0.6.3: Fixed telemetry — simple GET URL, no JSON escaping. Gene Map receives real data.
v0.6.2
v0.6.2: Fix exec commands — split into two simple calls for reliable execution
v0.6.1
v0.6.1: Default telemetry — every protocol trigger now feeds the Gene Map automatically. Anonymous event only, no user data.
v0.6.0
v0.6.0: Protocol 8 — Agent Role Guard. 8 total protocols. Hook v0.3 with role enforcement. Zero Clawdi API dependency.
v0.5.0
v0.5.0: Protocol 7 — Behavioral Failure Detection (silent abandonment, unverified completion via sessions_spawn, instruction drift). Hook v0.2 with zero Clawdi API dependency.
v0.4.1
- Improved error classification: now use a simple table to map error signals to standard codes before querying the Gene Map.
- Clarified repair strategy selection: clearly specify confidence thresholds and fallback behavior for protocol use.
- Streamlined Gene Map query instructions for error handling.
- No changes to protocols, privacy, or logging—core logic remains the same.
v0.4.0
v0.4.0: SKILL.md now queries Gene Map API before applying protocols. Strategies backed by real repair data — confidence scores improve as more agents use Vial.
v0.3.0
v0.3.0: Addressed security scan concerns — softer language, declared telemetry endpoint, scoped re-auth to OpenClaw connector flows only
v0.2.0
Vial Self-Healing Runtime v0.2 is a significant update focused on proactive, protocol-driven error recovery for AI agent reliability.
- Introduces mandatory checks before and after all external tool calls to detect auth issues and log failures.
- Adds clear, protocol-based actions for handling loops, step chains, rate limits, auth errors, session problems, and silent failures.
- Specifies logging standards for each protocol to /tmp/vial.log for centralized diagnostics.
- Requires session startup checks, with recent repair logs reported to users.
- Provides tailored instructions and best practices for specific agent types (CEO, Crypto, Stock).
- Emphasizes proactive detection and repair—minimizing silent failures and improving transparency.
Metadata
Frequently Asked Questions
What is Openclaw?
Vial self-healing runtime — 8 protocols covering technical failures, behavioral failures, and agent role enforcement. It is an AI Agent Skill for Claude Code / OpenClaw, with 114 downloads so far.
How do I install Openclaw?
Run "/install vial-self-healing" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Openclaw free?
Yes, Openclaw is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Openclaw support?
Openclaw is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Openclaw?
It is built and maintained by adrianhihi (@adrianhihi); the current version is v0.6.3.
More Skills