Run engagement

Autonomous, validated, false-positive-filtered.

Targets (one URL per line — multi-target)

Backend (runtime)

Provider

Model

OOB collaborator (optional)

Verbosity

Reinforcement learning Playwright MCP + screenshots

Agent library

…

Dropped into agents_md/vulns/ and orchestrated by the main agent on the next run.

Name (slug)

Title

Tests for

Severity

CWE

Methodology (markdown bullets)

System prompt (anti-false-positive rule)

Agent outputs & reinforcement-learning weights.

Run an engagement to populate.

Higher = historically more productive. Updated after every run.

No RL history yet.

PDF + HTML, generated via Typst.

No reports yet — run an engagement.

Execution mode, provider keys, orchestrator.

Execution mode

Main orchestrator agent

Default verbosity

Stored in memory for this session (only key presence is persisted to disk). Enables “via API” model use.