- name
- osint-overview
- description
- >
# OSINT-Only Operator Skill Catalog
This catalog is **passive**. No packets reach the target. Sandbox network
policy must restrict outbound to known-OSINT endpoints only (Shodan,
Censys, Hunter, GitHub API, crt.sh, Wayback, etc.).
## Playbooks
> **Inline technique reference — not separately loadable skills.** The entries below
> are summarized here for direct use; there is no separate `SKILL.md` to open for
> each. Do NOT call the skill loader on them — apply the technique with your tools
> using this summary and the Workflow in this file.
| Technique | Use for |
|---|---|
| **domain-pivots** | Whois history, reverse-IP, related-domain enumeration |
| **ct-logs** | crt.sh / Censys cert search for subdomain enumeration |
| **shodan-fingerprint** | Shodan host search; service / banner / ssl.cn pivots |
| **censys-pivots** | Censys cert/host/services pivots |
| **github-code-search** | GitHub code search for org's leaked secrets / config |
| **wayback-archives** | Wayback Machine API; retired endpoints, deleted docs |
| **breach-data** | HIBP / DeHashed (RoE-permitted only); credential reuse paths |
| **employee-profiling** | LinkedIn search (Sales Nav / manual), email-format inference |
| **asn-bgp** | ASN ownership, BGP table snapshots, RIR records |
| **maltego** | Maltego CLI graph projection; transform chain |
| **cryptocurrency** | Chain analysis (Etherscan / Mempool.space / Arkham) for crypto-adjacent targets |
| **geospatial** | Image geolocation, EXIF mining, satellite/streetview cross-reference |
## Workflow
1. **Seed**: from the engagement target (domain, company name, brand).
2. **Domain layer**: whois, reverse-IP, CT logs → enumerate every
subdomain and adjacent domain.
3. **Service layer**: Shodan + Censys against discovered IPs → service
inventory (NO probing; just consume cached scan data).
4. **Code layer**: GitHub code search for the target's org name, domain
names, internal package names, AWS account IDs.
5. **People layer**: employees via LinkedIn; email format inference;
HaveIBeenPwned for credential reuse.
6. **Infrastructure layer**: BGP + ASN ownership; Wayback retired
endpoints; SSL/TLS cert history.
7. **Synthesis**: project the graph into Neo4j as a pre-engagement map;
hand off to the Recon agent for active confirmation only if RoE
permits.
## Network policy
```
[osint-operator container] → outbound to: shodan.io, api.censys.io,
api.hunter.io, api.github.com,
crt.sh, archive.org, hibp/api/v3,
maltego.com, etherscan.io, ...
NO outbound to the engagement target.
```
The sandbox-net policy for OSINT engagements pins this allowlist. Any
attempted egress to the actual target IP/domain triggers a SafeCommand
refusal.
## Why split from Recon
Recon is active by default — port scans, version probing, directory
brute-forcing. Bug-bounty programs and pre-engagement scoping work
explicitly forbid touching production. OSINT-only enforces the
no-touch contract structurally rather than relying on the agent prompt
to remember.
Voir sur GitHub