Skip to content
Back to skills

Automation Recon Pipeline

ASecurity

Chain recon tools into a repeatable, resumable pipeline for continuous bug-bounty coverage. Load on "automate recon", "recon pipeline", monitoring many programs, or scaling subdomain→ live→scan. Signals: wildcard scope at scale, wanting scheduled/continuous discovery.

  • 21 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 11, 2026
ai-agentsgogit

Security analysis

A100/100

Scanned October 11, 2026

npx -y skills add NoorQureshi/security-skills --skill automation-recon-pipeline --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Automation Recon Pipeline?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Automation Recon Pipeline
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/noorqureshi-automation-recon-pipeline-security-skills/badge)](https://www.skillsdirectory.com/skills/noorqureshi-automation-recon-pipeline-security-skills)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: automation-recon-pipeline
description: >
  Chain recon tools into a repeatable, resumable pipeline for continuous bug-bounty coverage.
  Load on "automate recon", "recon pipeline", monitoring many programs, or scaling subdomain→
  live→scan. Signals: wildcard scope at scale, wanting scheduled/continuous discovery.
domain: automation
type: technique
stability: learning
modes: [bugbounty]
severity: info
mitre: [T1595]
tools: [subfinder, dnsx, httpx, naabu, nuclei, notify, anew]
schema_version: 1
---

# Recon automation pipeline

## When it applies
You track one or more wildcard programs and want discovery to run repeatably (and on a
schedule), surfacing only *new* assets/findings instead of re-reviewing everything.

## Why it works
Recon is a directed flow: roots → subdomains → resolve → live → ports → nuclei. Wiring the
ProjectDiscovery tools together (each reads/writes lines) makes it composable, and diffing
against last run (`anew`) turns it into continuous monitoring that alerts on change.

## Method
1. **Pipeline (each stage feeds the next):**
   ```
   subfinder -dL roots.txt -all -silent \
     | dnsx -silent -a -resp-only \
     | httpx -silent -sc -title -tech-detect \
     | tee live.txt \
     | nuclei -silent -severity low,medium,high,critical
   ```
2. **Diff for "only new"**: pipe each stage through `anew subs.txt` / `anew live.txt` so reruns
   emit only newly-seen lines — the basis of continuous monitoring.
3. **Notify**: pipe results to `notify` (Slack/Discord/Telegram) so new hosts/findings alert you.
4. **Schedule** with cron/systemd-timer/GitHub Actions; persist state files per program so runs resume.
5. **Keep templates current**: `nuclei -update-templates` before each run.

## Gotchas
- Stay in scope: feed only in-scope roots, filter out-of-scope hosts before scanning.
- Respect rate limits (`-rl`, `-c`) and program automation rules — don't hammer.
- De-dupe wildcard DNS (`dnsx` wildcard filtering) or you'll alert on noise forever.

## Verify success
A rerun surfaces only newly-appeared assets/findings and notifies you — hands-off continuous coverage.

## References
ProjectDiscovery pipeline docs; TomNomNom `anew`; nuclei templates.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.