Sightglass instruments AI coding agents to capture dependency choices, surfacing risks and biases in the agent's decision-making process.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install sightglass
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install sightglass using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Sightglass?
Sightglass is a powerful instrumentation layer designed to bring transparency to the software supply chain when working with AI agents. While human developers usually leave a trail of reasoning for their choices, AI agents often operate on invisible training data biases, selecting packages based on outdated popularity or training recall rather than technical merit. By integrating Sightglass into your Openclaw Skills, you gain the ability to monitor every tool selection and architectural choice in real-time.
This tool is essential for developers who want to ensure their AI-assisted projects remain secure and optimized. It classifies how an agent discovers a dependency—whether it was a proactive search or a simple hallucination from memory—allowing you to maintain high standards for your codebase. Using Sightglass within your Openclaw Skills ecosystem transforms a black-box automation process into a verifiable, audited development workflow.
Sightglass Use Cases
- Auditing automated pull requests to ensure AI agents aren't introducing unmaintained or vulnerable dependencies.
- Identifying training data bias in Openclaw Skills where agents favor older libraries over modern, more efficient alternatives.
- Monitoring background agent sessions to maintain a clear audit trail of all architectural changes.
- Comparing agent reasoning across different LLMs to determine which models make the most informed dependency choices.
How Sightglass Works
- The setup script initializes the environment by installing the @sightglass/cli and configuring the watcher daemon.
- Users authenticate with the cloud dashboard to enable persistent session history and advanced risk analysis.
- A background watcher daemon is started to monitor file system changes, package manager calls, and agent tool executions.
- Integration hooks wrap the agent session, recording the project context before execution and performing a deep analysis immediately after.
- The system generates a session summary that highlights training recall percentages, identified risks, and missed alternatives.
Sightglass Setup
To begin using this skill, execute the setup script to install the necessary CLI components and verify the watcher daemon:
./skills/sightglass/setup.sh
Log in to enable cloud analysis features:
sightglass login
Start the monitoring service before triggering your Openclaw Skills:
sightglass watch
Sightglass Data Schema & Taxonomy
Sightglass maintains a structured taxonomy of agent behavior and dependency metadata to provide actionable insights for Openclaw Skills users:
| Data Point | Description |
|---|---|
| Dependency Inventory | A full list of packages added, removed, or upgraded during the agent session. |
| Discovery Classification | Categorizes the choice as Training Recall, Context Inheritance, Search-based, or User Directed. |
| Risk Level | Assessment of the dependency based on maintenance status and known vulnerabilities. |
| Bias Indicators | Analysis of whether the agent relied on memory or active research for its decisions. |
| Alternatives Report | A list of superior or safer packages the agent failed to consider. |
Sightglass Advanced Features
- Automatic session tracking via pre-spawn and post-session hooks tailored for Openclaw Skills.
- Real-time watcher daemon that captures low-level tool calls and dependency installations as they happen.
- Multi-format analysis output including JSON and Markdown for easy integration into CI/CD pipelines.
- Cloud synchronization with sightglass.dev for cross-project visibility and historical trend analysis.
- Granular time-window analysis using relative flags to audit specific agent interactions.
SKILL.md
Loading
Related Openclaw Skills
Codifica
A file-based protocol for coordinating work between humans and AI agents using plain text files and Git as the source of truth.
GatewayStack Governance
A deny-by-default security and governance framework that intercepts every tool call to ensure agent safety and compliance.
GOYFILES External Bot Access
The core integration contract for external bots to securely onboard and interact with GOYFILES datasets and graph search tools.
Radio Copilot
A satellite pass orchestrator that sends real-time dish alignment data and WhatsApp alerts for manual or automated SDR reception.
NoticiasCangrejo
A high-efficiency news aggregation skill that fetches, ranks, and summarizes GNews articles into structured Markdown reports.
ByBit Order Book Backtester
A comprehensive end-to-end pipeline for downloading ByBit L2 order book data and simulating high-frequency trading strategies.
Featured*
