Nuclei sends targets requests described by YAML templates and reports the responses that match an exposure, a misconfiguration or a known vulnerability. It is a scanner: it produces findings to verify, not proof. The options below match the current CLI; confirm them with nuclei -h on your version. Scan only the scope agreed using the penetration test scoping checklist, and replace https://target.example with your target. To spot the exposed services first, see the Nmap cheatsheet.
Update and explore templates
# Nuclei version, then the installed template version
nuclei -version
nuclei -templates-version
# Update the templates
nuclei -update-templates
# List available tags, then the templates matching a filter
nuclei -tgl
nuclei -tl -tags exposure -severity medium,high,critical
Templates are written by the community and change often. Record the Nuclei and template versions in the report so the scan is reproducible. Read a template (-t path/template.yaml) before using it on a client: what it sends must be acceptable in production.
Run a first scan
# One target
nuclei -u https://target.example
# A list of targets (one per line)
nuclei -l targets.txt
# Findings only, without the banner or statistics
nuclei -u https://target.example -silent
# Header or cookie applied to all HTTP requests (test account)
nuclei -u https://target.example -H "Cookie: session=VALUE"
The first pass should stay limited: a minimum severity, a few tags, a reduced rate. An unfiltered scan launches thousands of requests.
Select templates
# By severity (info, low, medium, high, critical, unknown)
nuclei -u https://target.example -severity high,critical
# By tags
nuclei -u https://target.example -tags exposure,misconfig,tech
# By identifier (wildcards allowed)
nuclei -u https://target.example -id "git-config,CVE-2021-*"
# A specific template directory or file
nuclei -u https://target.example -t http/exposures/
# Automatic technology detection, then the associated templates
nuclei -u https://target.example -automatic-scan
| Option |
Purpose |
-tags / -etags (-exclude-tags) |
include or exclude by tag |
-severity / -exclude-severity |
filter by severity |
-id / -exclude-id |
include or exclude by identifier |
-t |
template file or directory |
-exclude-type |
exclude a protocol (dns, http, headless, tcp…) |
-new-templates |
only the templates from the latest release |
The long form -exclude-tags and the short form -etags are the same option. Always check the resulting list with -tl before starting.
Tune rate and load
# At most 20 requests per second, 10 templates in parallel, 10 hosts in parallel
nuclei -u https://target.example -rate-limit 20 -concurrency 10 -bulk-size 10
# Timeouts and retries
nuclei -u https://target.example -timeout 10 -retries 1 -rate-limit 20
By default, -rate-limit is 150 requests per second and -concurrency and -bulk-size are 25: that is too much for many production applications. Agree a rate with the client, watch for 429 and 5xx codes, and stop the scan if the service degrades. -max-host-error (30 by default) abandons a host that keeps producing errors; do not disable it (-no-mhe) on a fragile service.
Leave out intrusive templates
# Exclude risky template families
nuclei -u https://target.example -exclude-tags dos,intrusive,fuzz,default-login
# Exclude protocols or specific templates
nuclei -u https://target.example -exclude-type headless -exclude-id "identifier-to-skip"
# No external OAST server (interactsh): excludes templates that depend on it
nuclei -u https://target.example -no-interactsh
Tag names evolve: list them with -tgl and check with -tl what will really run. A tag such as dos or intrusive flags templates that may disturb a service, and default-login marks templates that try default credentials (authentication attempts, lockout risk); the absence of a tag does not guarantee they are harmless. DAST templates (-dast) send test payloads into parameters, and -headless templates drive a browser: enable neither without explicit agreement, and ideally try them on a staging environment first. By default, some templates use a public OAST server: check that the agreement with the client allows it, otherwise use -no-interactsh.
Save the results
# JSONL (one JSON line per finding) into a file
nuclei -u https://target.example -jsonl -output results.jsonl
# Markdown reports exported into a directory
nuclei -u https://target.example -markdown-export nuclei-report
# See the requests and responses exchanged (diagnosis)
nuclei -u https://target.example -id git-config -debug
# Send the traffic through Burp
nuclei -u https://target.example -proxy http://127.0.0.1:8080
JSONL is easy to process with jq or a spreadsheet and can be imported into a tracking tool. Keep it with the exact command, the date and the versions. Going through a test proxy helps to re-read the exchanges: see the Burp Suite cheatsheet.
Handle false positives
Every Nuclei finding is a hypothesis. Before reporting it:
- Read the template: what exactly does it test, and which
matcher fired?
- Replay the request by hand (
-debug, then curl or Repeater) and read the full response.
- Check the context: a custom error page or a WAF may return a
200 for everything.
- Confirm the component’s real version or configuration from another source.
- Durably set aside a noisy template with
-exclude-id, and note it in the report.
info findings (detected technologies, headers) help map the surface but are not vulnerabilities. Rate and prioritise the rest with the client, without taking the template’s severity as it stands.