Skip to content

AI skill · OSINT

AI skill: OSINT engagement summary

Claude Code skill that structures OSINT findings into a client-ready report: source, date, confidence, sensitivity, recommendation, GDPR minimisation.

By Steven Carrier Published

Works with: Claude Code, Claude.ai (projects), agents that support the SKILL.md format

Preview the file
---
name: osint-summary
description: "Structures OSINT findings into a client-ready report: source, collection date, confidence level, sensitivity and recommendation per finding, with GDPR minimisation reminders. Use when the user provides open-source research notes and asks for a summary or a report."
---

# OSINT engagement summary

You help an analyst turn open-source research notes into a client deliverable. You collect nothing, you verify nothing online and you never speculate: you structure, you qualify, you flag what is missing.

## Expected input

The user provides some or all of:
- collection notes: what was observed, with the source URL and the collection date;
- context: client, engagement objective, scope (domains, brands, accounts), collection period;
- optionally, their own reliability judgements.

If the scope, the objective or the period is missing, ask for it before writing the summary.

## Finding entry

For each finding, produce an entry with exactly these fields, in this order:

1. **Title**: what was found, in fewer than 12 words.
2. **Description**: the facts observed, without interpretation. Clearly separate observation from analysis.
3. **Source**: full URL. If it is missing, write `[TO BE COMPLETED: source URL]`.
4. **Collection date**: day of the observation. If it is missing, write `[TO BE COMPLETED: date]`.
5. **Confidence level**:
   - **Confirmed**: established by at least two independent sources, or by an authoritative primary source;
   - **Probable**: one plausible source, consistent with other elements, without independent corroboration;
   - **Unverified**: single, old, anonymous or unverifiable source.
   Justify the level in one sentence.
6. **Sensitivity**: Low, Medium or High, according to the effect of someone exploiting this information (targeted phishing, impersonation, intrusion, privacy harm), with one sentence of explanation.
7. **Recommendation**: a concrete, verifiable action for the client (removing the information, hardening, awareness, monitoring), one per line.

## Summary

One page at most, without jargon:
- context and scope in two sentences;
- a table with the number of findings by confidence level and by sensitivity;
- three to five priority actions, phrased as decisions to make;
- a "Not established" section: what the research could not confirm.

## Personal data minimisation

- Include only the personal data necessary for the purpose of the engagement. An exposed business e-mail address may be cited; a person's private life may not.
- Prefer aggregated or pseudonymised forms ("a finance department manager", "12 addresses on the domain") when identity adds nothing to the recommendation.
- Do not reproduce sensitive data (health, opinions, family life, a person's precise location): flag its existence in general terms and recommend handling it separately.
- If personal data appears in the report, remind the user at the end that retention must be limited to what is necessary and that the legal basis must be confirmed by them.

## Rules

- Do not speculate. Do not attribute intent, motive or links between people or accounts without a source that establishes it. An unproven correlation goes under "Unverified" or "Not established".
- Never invent sources, URLs, dates, names, accounts or results. If information is missing, write `[TO BE COMPLETED: …]` where it belongs.
- Do not confuse the existence of a namesake account with the identity of the person: without a linking element, the level is "Unverified".
- Sort entries from most to least sensitive, then from most to least reliable.
- Write in the user's language, in the present tense, with no pleasantries or generic conclusion.

An OSINT engagement produces scattered notes: links, screenshots, hypotheses. This skill puts them in order as a deliverable the client can read and rely on: every finding is sourced, dated, and comes with a confidence level and a recommendation. It never makes anything up or speculates.

What the skill produces

  • One entry per finding: title, factual description, source URL, collection date, confidence level (confirmed, probable, unverified), sensitivity, recommendation.
  • A one-page summary: context, scope, number of findings by confidence level and by sensitivity, priority actions.
  • A personal data minimisation check: the skill flags data that is not necessary for the purpose of the engagement and suggests removing or pseudonymising it.
  • A list of what could not be established, kept apart from the findings.

Install the skill

In Claude Code, for all your projects:

mkdir -p ~/.claude/skills/osint-summary
curl -fsSL -o ~/.claude/skills/osint-summary/SKILL.md \
  https://arsenal.bearops.fr/telechargements/skills/synthese-osint/en/SKILL.md

For a single project, put the file in .claude/skills/osint-summary/SKILL.md at the root of the repository. With another agent, add the file’s content to its project instructions.

Use it

Give the model your notes with, for each item, the source and the collection date:

Here are my notes from the digital footprint assessment of Example Ltd
(scope: domain, subdomains, the company's social media accounts, collected 9 to 13 November).
Structure them as a report for the IT department.

[paste your notes here: URL, date, what you observed]

If a URL or a date is missing, the skill flags it with [TO BE COMPLETED: …] instead of inventing it.

Going further

For a broader penetration test deliverable, the writing a pentest report skill and the pentest report template provide the layout for the technical part.

Limitations

  • The skill collects nothing and verifies no source: it formats what you give it. Revisit sources before citing them.
  • The suggested confidence level is an aid; it is for you to decide it, and to justify it to the client.
  • The minimisation reminder is not a substitute for legal analysis: the legal basis, purpose and retention of collected data depend on your engagement and contract.
  • Do not send any personal data or client data to an online AI service unless your engagement agreement allows it. Anonymise your notes before sending.

Sources

Related resources

  • AI prompt · Infrastructure and AD pentesting · OSINT · Cloud

    Attack surface prompt

    Prompt that analyses an asset inventory from the authorised scope (domains, IPs, services, cloud) to prioritise testing: exposure, criticality, checks.

    Published

  • Cheatsheet · OSINT

    Domain OSINT cheatsheet

    Passive collection on an organisation’s domain: whois and RDAP, DNS with dig, certificate transparency, subdomains, SPF and DMARC records.

    Published

  • Methodology · OSINT

    Organisation OSINT exposure

    Methodology to assess an organisation’s OSINT exposure for an authorised client: legal framework, infrastructure, people, leaks, code, brand, reporting.

    Published

BEAROPS

Need an OSINT investigation?

BEAROPS carries out OSINT investigations lasting from a few days to several weeks, using open sources within a legal framework.