SPECTRA FIELD MANUAL
EN/IT
Getting Started

A complete cybersecurity team — instantly available in your AI IDE.

Getting Started

SPECTRA installs as a set of agent skills into your AI coding environment (Claude Code natively, plus a Codex repo-native adapter). One command:

npx spectra-method install

Pick the modules an engagement needs — Core is always installed; add RTK, SOC, IRT, GRC as required.

Quickstart in 5 minutes

To see the whole flow before writing any YAML, scaffold the shipped demo — a safe, loopback-only engagement with sample findings and a worked War Room debrief:

spectra quickstart

It prints a guided tour: validate the engagement, preview a gated scan (127.0.0.1 plans, 8.8.8.8 is refused out of scope), snapshot the posture, export the findings, and open the War Room debrief. Nothing offensive runs — the demo targets loopback only.

Prefer a head start for a real engagement? Scaffold a scenario template instead:

spectra quickstart list                          # demo, web-pentest, cloud-ir, ot-assessment
spectra quickstart --template web-pentest --dest ./my-engagement

Each template carries the recommended agents and workflows for that scenario; fill the placeholders, set status: active, and validate.

1. Create an engagement

Everything in SPECTRA is bound to an authorized engagement: scope, Rules of Engagement, and authorization context. Start one:

/spectra-new-engagement

You’ll choose a lead persona first (Red, Blue, Incident, GRC, or a quick solo operator), then define scope, RoE and an optional noise budget. Nothing offensive runs until the engagement is created and gated.

2. Do the work

Each agent does the work, it does not just plan it. Talk to a persona by name — Viper (Red lead), Commander (SOC), Dispatch (incident handler) — or run a structured workflow:

/spectra-external-recon
/spectra-alert-triage
/spectra-incident-handling

3. Prove it

Run a reproducible benchmark, generate an attack-path graph, and produce the report — the report is the deliverable.

spectra-cli ...        # deterministic execution layer

Next

  • Architecture — the three layers.
  • Modules — what each kit contains.
  • Safety — the boundary that makes SPECTRA defensible.