Agent Infrastructure · Curated marketplace
install-openviking-memory
Install and configure the OpenViking long-term memory plugin for OpenClaw via natural conversation.
Composite
C 4.1 · A 0.0
How we got there
Our evaluation
Tier-2 Review: install-openviking-memory
What we attempted
We pulled install-openviking-memory into a clean container and tried to exercise it end-to-end: satisfy the declared prerequisites, trigger the skill via its documented natural-language phrases, run the install path, and smoke-test a minimal invocation. The skill advertises itself as an agent-driven installer for the OpenViking long-term memory plugin on OpenClaw — "auto-capture + auto-recall, cross-session" — with a plugin-manager install path, an ov-install backup path, wizard configuration, slot activation, gateway restart, verification, and uninstall.
What failed
Nothing passed. The harness recorded 0 passed, 1 partial, 0 failed, 1 skipped. That is a clean sweep of non-completion, and it is worth being precise about why, because the failure is structural rather than a bug in a command.
Two hard dependencies gate the entire skill:
- An OpenClaw host with a plugin manager. The skill is a conversational agent skill, not a standalone CLI package. Installation is performed through OpenClaw's plugin manager, with
ov-installexplicitly demoted to a backup path. Without an OpenClaw runtime, there is no install surface to invoke. - A pre-running OpenViking server. The SKILL.md states plainly that it "assumes the OpenViking server is already running." No server, no target.
Neither is available in a clean container, and the skill does not provide a path to provision them — it assumes them.
What we observed
- install (partial): The prerequisites check and the server-availability assumption are stated, but the truncated SKILL.md contains no concrete install command string. The instruction is essentially "the agent runs everything," which is fine for a hosted agent but leaves a harness with nothing deterministic to execute. We could confirm the shape of the install flow (plugin manager first,
ov-installas fallback) but not run it. - smoke-invocation (skipped): Triggers are natural-language phrases ("install OpenViking", "set up memory", "接入 OpenViking", etc.). With no OpenClaw runtime and no running server, a minimal invocation cannot be meaningfully executed. No README invocation example appears in the provided content.
The failure mode is therefore environmental and under-specified, not a logic error we can point to. We cannot distinguish "skill is broken" from "skill is fine but unrunnable here."
Rating caveat
The composite of 4.1 / 5.0 (D1 5.0, D2 4.0, D3 4.0, D4 4.5, D5 2.5) should be read as theoretical. It reflects the quality of the written skill — trigger clarity and self-containment genuinely look strong — but it is not corroborated by execution. Until a physical re-run on a real OpenClaw host against a live OpenViking server resolves the two blockers, the score is an assessment of documentation, not behavior. D5 (reusability, 2.5) is the dimension most likely to move once tested, and it may move down: a skill that cannot be exercised outside one specific host is, by construction, low-reusability.
Does it still seem valuable in principle?
Yes — conditionally. The problem it solves is real: giving an agent persistent cross-session memory without the user touching a CLI is a legitimate, high-value capability, and the trigger surface is unusually well-designed (bilingual, intent-based, no command knowledge required). The plugin-manager-first / ov-install-fallback structure is sensible.
But its value is host-locked. It is only as reusable as OpenClaw + a running OpenViking server, and its SKILL.md would benefit enormously from (a) at least one concrete, copy-pasteable install/invocation example, and (b) an explicit "if the server is not running, do X" branch that a harness could actually follow. As written, it is a good skill for one ecosystem and an untestable one everywhere else. We recommend re-running on a provisioned host before treating the 4.1 as earned.
What we tried
Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-09-24.
Overall: broken. 0 tests passed, 1 partial, 0 failed, 1 skipped; key blocker: the skill requires an OpenClaw host and a pre-running OpenViking server, neither of which is available in a clean container, and no concrete install/invocation commands are specified in the SKILL.md.
Inferred dependencies: OpenClaw runtime with plugin manager, running OpenViking server, ov-install (backup path only).
| Test | Status | Notes |
|---|---|---|
| install | partial | The SKILL.md is a conversational agent skill, not a standalone CLI package; installation is performed through OpenClaw's plugin manager (with ov-install only as a backup path), so a generic 'install command' cannot be fully exercised without an OpenClaw host and a running OpenViking server. The prerequisites check and server-availability assumption are stated, but no concrete install command string is provided in the truncated content. |
| smoke-invocation | skipped | The skill triggers on natural-language phrases (e.g. 'install OpenViking', 'set up memory') and assumes the OpenViking server is already running; without that server and an OpenClaw runtime, a minimal invocation cannot be meaningfully executed. No README invocation example is present in the provided SKILL.md content. |
1 source verified
- Best source
skillsmp.com - Authority tier Tier 2 — Curated marketplace
- Stars ★ 24,010
- Source link https://skillsmp.com/skills/volcengine-openviking-examples-openclaw-plugin-skills-install-openviking-memory-skill-md ↗
- First published 2026-05-24
- Last modified 2026-09-24
Use this skill
/plugin install install-openviking-memory Head-to-head pages featuring install-openviking-memory
More in Agent Infrastructure
skill-creator
Create, edit, improve, or audit AgentSkills.
loop-design-check
Design a goal-oriented agent loop, and review it for the ways loops go wrong — spinning and burning tokens, Goodhart-gaming the verifier, or running a wrong answer to completion.
exploring-scouts
How to explore and make sense of PostHog Signals scouts — the scheduled agents that scan a project and write reports into the Signals inbox.
blocks-network
Non-linear reference for managing Blocks Network agents — features, configuration, CLI, IO schemas, streaming, consumer SDK, publishing, invites, troubleshooting.