← 返回 Skills 市场

Assistant Reliability Watchtower

Name: Assistant Reliability Watchtower
Author: wit-oc

作者 wit-oc · GitHub ↗ · v0.1.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install assistant-reliability-watchtower

功能描述

Deterministic reliability monitoring for OpenClaw assistant workflows. Use when you need to run ARW smoke probes, generate a daily digest, validate scorecard...

使用说明 (SKILL.md)

Assistant Reliability Watchtower

Run ARW through the repo-backed wrapper instead of rediscovering commands by hand.

Quick start

Use scripts/arw_skill.py from this skill for the common flows:

python3 scripts/arw_skill.py smoke
python3 scripts/arw_skill.py digest
python3 scripts/arw_skill.py verify-scorecard-evidence
python3 scripts/arw_skill.py verify-scorecard-preamble

The wrapper accepts --repo-root (or ARW_REPO_ROOT) so repo location stays explicit outside the skill contract, and it exports PYTHONPATH correctly.

Workflow

Read references/release-contract.md for scope, release posture, and anti-goals.
Read references/probe-catalog.md when you need to explain what ARW currently watches.
Read references/config-contract.md when you need to set thresholds, artifact location, or delivery evidence values.
Point the wrapper at the checked-out ARW repo with --repo-root /path/to/arw-watchtower when autodetection is not enough.
Use scripts/arw_skill.py for the common operational paths.
Keep writes inside the configured ARW repo and artifacts/arw/**.

Common tasks

Run smoke probes

python3 scripts/arw_skill.py smoke

Generate a digest

python3 scripts/arw_skill.py digest --window-hours 24

Run the validation evidence drill

python3 scripts/arw_skill.py verify-scorecard-evidence

Render the final operator preamble

python3 scripts/arw_skill.py verify-scorecard-preamble

Notes

This RC surface is intentionally repo-backed for now. It is an internal release candidate, not a marketplace-polished public release.
verify-scorecard-evidence and verify-scorecard-preamble now use skill-level dry-run delivery values instead of inheriting repo-local recipient ids.
Prefer RC1 work that improves packaging, operator clarity, portability, or high-value probe coverage.
Avoid expanding the backlog with new micro-gates unless they directly unblock RC1.

安全使用建议

This skill is coherent with its description: it’s a thin wrapper that runs ARW code in a repository you point it at. Before using it, verify the ARW checkout is trusted: review arw/run.py and arw.alert (and any modules imported by them), and run the skill in an isolated environment if the repo is untrusted. Check asset/example-config.json so delivery fields default to dry-run (avoid accidentally sending to real recipients), and prefer passing --repo-root explicitly rather than relying on autodiscovery. There are no required credentials in the skill itself, but the wrapper will execute repository code (both in-process import and via subprocess -m), which is the main operational risk.

功能分析

Type: OpenClaw Skill Name: assistant-reliability-watchtower Version: 0.1.0 The skill acts as a wrapper for an external repository ('ARW'), utilizing high-risk execution patterns in `scripts/arw_skill.py` such as `subprocess.run` to execute system commands and dynamic Python imports via `sys.path` manipulation. While these capabilities are aligned with the stated purpose of managing a repo-backed watchtower, the reliance on environment variables (`ARW_REPO_ROOT`) and CLI arguments to define execution paths creates a risk of arbitrary code execution if the environment or agent instructions are manipulated. No evidence of intentional malice or data exfiltration was found, but the architectural reliance on executing external, unverified code warrants a suspicious classification.

能力评估

✓ Purpose & Capability

Name/description match what the package actually does: a repo-backed wrapper that runs arw.run (smoke, digest, validation flows), manages PYTHONPATH, and writes artifacts under artifacts/arw. No unrelated binaries or credentials are requested.

ℹ Instruction Scope

SKILL.md and the script limit behavior to running ARW flows in a checked-out repo and reading the included docs/config. However, the wrapper spawns the repository's code (python -m arw.run) and imports repository modules (arw.alert) which will execute arbitrary code from that repo at runtime; this is expected for a repo-backed wrapper but is an important operational risk if the repo is untrusted.

✓ Install Mechanism

No install spec; the skill is instruction-plus-script only. Nothing is downloaded or written by an installer, minimizing supply-chain risk from the skill bundle itself.

✓ Credentials

No required env vars or credentials. The script optionally honors ARW_REPO_ROOT and sets PYTHONPATH to include the repo root; config delivery fields default to dry-run values. No secrets or unrelated platform credentials are requested.

ℹ Persistence & Privilege

always is false (normal). The skill runs subprocesses and imports code from the target repo—reasonable for its role but gives the skill the ability to execute arbitrary repo code and produce side effects in the repo working directory. Autonomous invocation is allowed by default (platform default) but not, by itself, a new risk here.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install assistant-reliability-watchtower
安装完成后，直接呼叫该 Skill 的名称或使用 /assistant-reliability-watchtower 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.1.0

Initial public RC1 release: packaged OpenClaw skill, repo-backed wrapper commands, and RC release/config/probe docs.

元数据

Slug assistant-reliability-watchtower

版本 0.1.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题