ShakerScan walkthrough

This guide follows the current first-run flow through an AI coding agent and the web UI. It uses text and live routes so it does not drift with historical screenshots.

1. Install and open ShakerScan

curl -fsSL https://install.shakerscan.com | sh

The installer starts the Docker stack on local interfaces. Open:

For remote access over Tailscale, use:

shakerscan start --remote
shakerscan status

Use the browser URLs printed by status; API commands running on the ShakerScan host still use http://localhost:8080.

2. Start an agent in the runtime

shakerscan agent codex
# or: shakerscan agent claude
# or: shakerscan agent opencode

Starting inside ~/.shakerscan lets the agent load AGENTS, CLAUDE, and the task-specific skills. Try:

Check whether ShakerScan is healthy.
Run a fast passive Scan on https://app.example.test.
Show active critical and high findings.

The agent checks health, submits work through ShakerScan, reports the returned ID and UI link, and stops instead of waiting for a long-running scan.

3. Run a first scan in the UI

  1. Open New Scan.
  2. Enter a target you own or are explicitly authorized to test.
  3. Keep the default Balanced coverage budget.
  4. Leave active testing off for the first run.
  5. Submit and open the returned Scan Detail page.

Scan Detail shows progress, current phase, logs, partial-result warnings, coverage, findings, proof state, deployment decision, and the final report.

The Thorough budget increases ceilings but does not grant active authority. Confirm authorization with the agent before enabling Active testing, and follow the warnings in the UI. Do not use active probes on third-party systems without explicit permission.

4. Review and verify findings

Open Findings to filter by source, severity, status, target, verification mode, or last-seen window. Finding Detail separates the reported issue from its evidence and verification history.

Use:

  • Retest to queue deterministic verification with optional AI escalation;
  • Evidence to inspect durable evidence objects and content-free exports;
  • status actions for resolved, false-positive, or accepted-risk triage;
  • Verify finding for a bounded deterministic proof attempt on an authorized, target-linked web finding.

A title, HTTP 200, reflection, or model opinion is not proof. Look for the server-derived proof state, deterministic replay, principal/control comparison, and linked evidence.

5. Choose the next workflow

GoalUI area
Maintain endpoint inventory and close coverage gapsContinuous ASM
Review assets, relationships, and attack pathsExposure
Test chat, RAG, agent, widget, or MCP endpointsAI Gate (preview)
Check a model artifact before deploymentModel Intake (preview)
Reproduce browser/auth workflows manuallyInteractive Testing
Review evidence cleanup or exportsEvidence
Follow scans, scheduled work, investigations, and exportsTimeline
Start an AI-driven investigation with bounded active testingHunt
Preview a natural-language operation without executing itAI Operations Router

The Campaigns page is a read-only mission-action ledger. Start an investigation from Hunt, then continue its keyless turn loop from the coding-agent session. The UI shows the current transcript, evidence-backed suspected findings, and deterministic verification results. Gated execution is enabled in standard installs, so no extra server setting or stored LLM key is needed; the target authorization and expiring approval prompts still apply.

6. Useful agent requests

Show ShakerScan status and use the correct remote UI URL.
Run a balanced passive Scan on my authorized staging application.
Explain the proof and coverage gaps for scan <id>.
Show active Hunt findings for this target.
What should Continuous ASM test next?
Run Hunt on this authorized registered target.
Check this model artifact without executing it.

See the README for installation, workflow selection, CLI, and troubleshooting. The Functionality Reference contains the exhaustive UI/API/CLI/skill map.

This page is rendered from WALKTHROUGH.md in the open-source repository at v2.5.4. When documentation and implementation disagree, the code, database schema, and tests at that tag are authoritative.