hypnguyen1209/offensive-claude / skills/recon-osint
recon-osint
VerifiedUse when mapping a target's external attack surface or gathering OSINT — subdomain enumeration, attack-surface mapping (httpx/katana/JS secrets), subdomain takeover, multi-cloud/Azure tenant recon, GitHub secret dorking, breach/infostealer credential intel, CVE prioritization (EPSS/KEV)
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.8 KB |
| references/attack-surface-mapping.md | markdown | 7.3 KB |
| references/breach-credential-intel.md | markdown | 8.1 KB |
| references/cloud-saas-recon.md | markdown | 8.3 KB |
| references/cve-exploit-intel.md | markdown | 8.1 KB |
| references/subdomain-discovery.md | markdown | 7.6 KB |
| references/subdomain-takeover.md | markdown | 7.5 KB |
| scripts/breach_intel.py | script | 9.9 KB |
| scripts/cloud_asset_enum.py | script | 8.6 KB |
| scripts/cve_prioritizer.py | script | 8.8 KB |
| scripts/hackerone_public_recon.py | script | 5.1 KB |
| scripts/js_secret_hunter.py | script | 7.4 KB |
| scripts/recon_orchestrator.py | script | 8.2 KB |
| scripts/subdomain_takeover.py | script | 8.4 KB |
| scripts/wordlist_ranker.py | script | 4.3 KB |
references/cve-exploit-intel.md
View on GitHub# CVE & Exploit Intelligence — Enrichment and Prioritization
Cluster: turning a technology fingerprint / exposure map into a *ranked* list of weaponizable CVEs
using authoritative feeds. ATT&CK: T1592.002 (Gather Victim Host Info: Software), T1596 (Search Open
Technical DBs), T1596.005 (Scan Databases). CWE-1395 (Dependency on Vulnerable Third-Party
Component); per-finding CWE is carried from the CVE record.
## Theory / Mechanism
A fingerprint (e.g. `nginx 1.18.0`, `Confluence 7.13`, `Citrix NetScaler 13.1`) maps to many CVEs,
but only a few are actually weaponizable against *this* target. Prioritization is **layered** — no
single score suffices:
- **CVSS** — potential severity, standardized communication. *Insufficient alone*: a notable share
of CVEs exploited in Q1-2025 carried only Medium base scores, so a CVSS≥7 filter misses real
in-the-wild threats.
- **EPSS** — ML probability the CVE is exploited in the wild within 30 days (0-1 + percentile).
Sorts large pools by near-term likelihood. Can lag a brand-new exploitation wave by a day or two.
- **CISA KEV** — human-verified *confirmed* active exploitation. First-tier, no model lag.
- **Exposure context** — is the vulnerable component internet-facing on the target? (Shodan
InternetDB, your httpx fingerprint, port map.) A KEV CVE on an exposed asset is top priority.
Consensus model: **rank by `(in KEV) → high EPSS → exposure → CVSS for ties`.**
## Modern 2024-2026 Facts (verified)
- **NVD CVE 2.0 API** is now the canonical source (legacy data feeds removed **20 Aug 2025**). It
embeds KEV fields (`cisaExploitAdd`, `cisaActionDue`, `cisaRequiredAction`). NVD now holds 360k+
records and paginates via `startIndex`/`resultsPerPage` (max 2000/page; rate-limit ~5 req/30s
without key, ~50 with).
- **The 2026 selective-enrichment shift (critical):** CVE submissions rose 263% (2020→2025); NVD
can't keep up. **From 15 Apr 2026** NVD *prioritizes* enrichment for: CVEs in CISA KEV (goal: 1
business day), CVEs for federal-government software, and EO-14028 "critical software." Everything
else still *enters* the NVD but many records get **no immediate CVSS/CPE enrichment**, and
pre-Mar-2026 backlog CVEs move to "Not Scheduled." => You must build your *own* enrichment stack;
don't assume a fresh CVE has a CVSS/CPE in NVD.
- **EPSS v4** (released **Mar 2025**) ingests 250k+ threat-intel data points/day, scores within 24h
of NVD publication, and **now scores backlog CVEs that lack an official CVSS** — closing the gap
the 2026 NVD change opens.
- **Shodan InternetDB** (`https://internetdb.shodan.io/<ip>`) is a *free, unauthenticated* endpoint
returning known CVEs, open ports, hostnames, and tags for an IP — the cheapest "exposure" layer.
- **GreyNoise** provides "# unique IPs scanning for CVE-X in last 90d" — a live exploitation-interest
signal usable as a tiebreaker.
## Complete Working Commands
### 1. Map fingerprint → candidate CVEs
```bash
# Offline ExploitDB:
searchsploit "Confluence 7.13"
searchsploit -j "nginx 1.18.0" | jq '.RESULTS_EXPLOIT[] | {Title, Path}'
# NVD 2.0 keyword search (use an API key to avoid throttling):
curl -s "https://services.nvd.nist.gov/rest/json/cves/2.0?keywordSearch=Confluence&resultsPerPage=20" \
-H "apiKey: $NVD_API_KEY" \
| jq -r '.vulnerabilities[].cve | [.id, (.metrics.cvssMetricV31[0].cvssData.baseScore // "n/a")] | @tsv'
# By CPE (precise version match):
curl -s "https://services.nvd.nist.gov/rest/json/cves/2.0?cpeName=cpe:2.3:a:atlassian:confluence:7.13.0:*:*:*:*:*:*:*" \
-H "apiKey: $NVD_API_KEY" | jq -r '.vulnerabilities[].cve.id'
```
### 2. Enrich each CVE: KEV + EPSS + GreyNoise
```bash
CVE=CVE-2023-22515
# CISA KEV (is it confirmed exploited?):
curl -s "https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json" \
| jq --arg c "$CVE" '.vulnerabilities[] | select(.cveID==$c) | {vendorProject, dateAdded, dueDate, knownRansomwareCampaignUse}'
# EPSS v4 score + percentile:
curl -s "https://api.first.org/data/v1/epss?cve=$CVE" \
| jq -r '.data[] | "\(.cve) epss=\(.epss) pct=\(.percentile)"'
# GreyNoise scanning interest (free RIOT/community where available):
curl -s "https://api.greynoise.io/v3/community/$CVE" 2>/dev/null
```
### 3. Exposure layer — Shodan InternetDB (free) + targeted Shodan
```bash
# Free, unauth, per-IP CVE+port map (great for bulk):
for ip in $(cat ips.txt); do
curl -s "https://internetdb.shodan.io/$ip" \
| jq -c --arg ip "$ip" '{ip:$ip, ports, vulns, hostnames, tags}'
done > internetdb.jsonl
# Authenticated Shodan for richer product/version + vuln view:
shodan host $IP
shodan search "hostname:$DOMAIN" --fields ip_str,port,product,version
```
### 4. One-shot ranked output (this skill's tool)
```bash
# Takes httpx fingerprints + IP list, queries NVD/EPSS/KEV/InternetDB, emits ranked JSONL:
python3 scripts/cve_prioritizer.py --from-httpx httpx_live.jsonl --ip-file ips.txt \
--nvd-key "$NVD_API_KEY" -o cve_ranked.jsonl
# Top priorities = KEV-listed AND internet-exposed AND high EPSS:
jq -r 'select(.in_kev and .exposed) | [.priority, .cve, .epss, .cvss, .product] | @tsv' \
cve_ranked.jsonl | sort -rn | head -20
```
### 5. Confirm exploitability (don't just pattern-match)
```bash
# Validate the actual presence/exploitability, not just a version string:
nuclei -u https://$TARGET -t http/cves/ -tags $CVE -severity high,critical -o nuclei_confirm.txt
# Find a real PoC:
curl -s "https://api.github.com/search/repositories?q=$CVE+poc&sort=updated&per_page=5" \
| jq -r '.items[] | "\(.html_url) ⭐\(.stargazers_count) upd:\(.updated_at)"'
```
## Prioritization Matrix (priority score the script assigns)
| in KEV | EPSS | Internet-exposed | CVSS | → Priority |
|--------|------|------------------|------|-----------|
| yes | any | yes | any | **P0 — exploit now** |
| yes | any | unknown/no | any | P1 |
| no | ≥0.5 | yes | ≥7 | P1 |
| no | ≥0.1 | yes | any | P2 |
| no | <0.1 | yes | ≥9 | P2 |
| no | <0.1 | no | any | P3 — defer |
## Detection
CVE enrichment touches only third-party APIs (no target telemetry). The *follow-on* nuclei
confirmation scan is the detectable part (see attack-surface-mapping.md). For defenders, the
mirror-image control is running the same KEV×EPSS×exposure stack against their own asset inventory:
```yaml
# Sigma — confirmation scan probing a specific KEV CVE path (web access logs)
title: Targeted KEV CVE Exploitation Probe
id: 4d2c9e7f-recon-cveprobe
status: experimental
logsource:
category: webserver
detection:
selection_path:
cs-uri-stem|contains:
- '/server-info' # e.g. CVE-2023-22515 Confluence probe
- '/setup/setupadministrator.action'
selection_ua:
cs-user-agent|contains: 'nuclei'
condition: selection_path or selection_ua
level: high
tags: [attack.t1190, attack.t1595.002]
```
## OPSEC
- **Touches:** NVD/EPSS/KEV/Shodan/GitHub APIs only — invisible to the target. InternetDB is
passive (Shodan already scanned the IP; you query Shodan, not the target).
- **Cleanup:** none.
- **Evasion / discipline:** map CVEs only to **in-scope** assets; the loud step is the nuclei
*confirmation* scan against the target — gate it behind ROE and rate-limit it. Cache NVD results
and honor the API rate limit (use a key) to avoid throttling/bans. Given the 2026 NVD enrichment
change, do not assume a missing CVSS/CPE means "not vulnerable" — fall back to EPSS + KEV + vendor
advisory.
## References
- NVD Vulnerability APIs (2.0) — https://nvd.nist.gov/developers/vulnerabilities ; legacy feeds removed 20 Aug 2025
- ReversingLabs, "Selective NVD enrichment: Why NIST's shift matters" (15 Apr 2026 prioritization) — https://www.reversinglabs.com/blog/nist-selective-nvd-enrichment
- FIRST EPSS v4 (Mar 2025) — https://www.first.org/epss/ ; decryptiondigest EPSS workflow guide — https://www.decryptiondigest.com/blog/epss-vulnerability-prioritization-patching-workflow-guide
- CISA KEV Catalog — https://www.cisa.gov/known-exploited-vulnerabilities-catalog ; Nucleus KEV enrichment guide — https://nucleussec.com/resources/guides/guide-to-cisa-kev-enrichment/
- Shodan InternetDB — https://internetdb.shodan.io/ ; runZero "KEVology" — https://www.runzero.com/resources/kevology/
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |