How to use the collector-hand-skill agent skill for OSINT

What the skill makes the agent do

The skill is prompt-only guidance. It describes methods and does not itself run code or fetch anything, so the agent still does the searching and reading using whatever it already has access to. What the skill adds is a consistent method: the agent plans its target and scope, collects raw data, processes it into entities and relationships, analyses the results, produces a report, and refines its queries based on what worked.

The skill also sets out a five-tier source reliability scale, from official and primary sources such as company filings and press releases, down to anonymous or unverified claims. Before data goes into a knowledge graph, the agent is meant to check it against this scale.

• Collection cycle: planning, collection, processing, analysis, dissemination and feedback
• Source tiers: official/primary (very high) down to anonymous/unverified (low)
• Search query patterns for market, business, competitor, person and technology research
• Entity and relation schemas for a simple knowledge graph
• Change detection between snapshots with critical, important and minor significance levels
• Optional sentiment tracking when track_sentiment is enabled

Install collector-hand-skill in Claude

Download the ZIP file for the skill from this page. The ZIP contains the skill files, which are HAND.toml and SKILL.md. You can install it in one of two ways.

For a personal install, unzip the file into your skills folder at ~/.claude/skills/. For a single project, unzip it into the .claude/skills/ folder inside that project. You can also upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills.

• Personal: unzip into ~/.claude/skills/
• Project only: unzip into .claude/skills/ inside the project folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

How to trigger the skill with natural requests

You do not need special syntax. Ask for research in plain language that matches the kind of work the skill covers: a company, a competitor, a person from a public role, or a technology. The agent should use the skill's methods when the request calls for structured collection, source rating or change tracking.

Good requests name the target, the scope and the output you want. Mention a time frame if you want changes since a previous check, because the skill's change detection depends on comparing against an earlier snapshot.

• Map the main competitors of [company] and summarise how their positioning differs, rating each source by reliability.
• Build a structured note on [company]: leadership, recent acquisitions, partnerships, and where each fact came from.
• Compare the current state of [company] against the snapshot from our last collection and flag leadership changes or acquisitions.
• Give me a market analysis of [industry] trends with sources ranked by tier.

Search query patterns the skill uses

The skill includes ready-made query patterns for each focus area, with placeholders such as [company] or [technology] that the agent fills in. Business queries, for example, combine a company name with terms like revenue, CEO, partnerships or annual report, and can use site: restrictions to target a specific domain.

Example business intelligence query patterns from the skill:

Understand the source tiers and significance levels

Source tiers tell you how much weight a claim deserves. Tier 1 covers company filings, government data and press releases. Tier 2 covers news agencies and research institutions. Tier 3 covers industry publications and analyst reports. Tier 4 covers forums, social media and review sites, and Tier 5 covers anonymous posts and rumours.

Changes are sorted by significance. The skill lists leadership changes, acquisitions and mergers, large funding rounds, product discontinuations and legal or regulatory issues as critical, meaning immediate alert. New launches, partnerships, pricing changes and major customer wins or losses are important. Blog posts, minor patches and conference appearances are minor.

Safety, limits and what to check first

The skill does not run scripts and does not use the network, and it lists no requirements. Any web searching depends on the tools your agent already has, not on this skill. The skill's own guidance also includes person-tracking query patterns, and those can enable stalking or harassment when used on private individuals. Check legal and ethical limits before researching any person.

The source-tier and sentiment scoring are heuristics, not validated methods. Do not rely on them alone for legal, financial or investment decisions. Verify key facts independently.

• No scripts are run and no network use is declared by the skill
• Do not use person-tracking patterns on private individuals
• Verify legal, financial and investment conclusions independently

Tips for better results

Give the agent a clear target and a fixed set of questions so that repeated runs stay comparable. Keep the previous snapshot and ask the agent to compare against it, since change detection only works when there is an earlier state to compare with. Ask for source tiers in the output so you can see which claims rest on primary evidence.

• Name the target, scope and output format in each request
• Keep earlier snapshots so changes can be detected
• Ask for each claim to be tagged with its source tier and confidence