dsh-pentest-bugtrace

Plugging a mature pentest engine into a dsh profile usually means hand-assembling an MCP bridge, writing a runbook, and tuning a persona prompt, each step prone to drift. dsh-pentest-bugtrace collapses that chain into a single plugin install: the result is a ready-to-use BugTraceAI pentest agent that insists on authorization first and verification before any finding is reported.

Three layers of capability stack together. At the base sit seven native MCP tools covering scan start, progress polling, finding queries, abort, report export, vulnerability explanation, and remediation suggestions; when the backend is unavailable the bridge fails open and falls back to the local REST endpoint or CLI command so the workflow never breaks. In the middle ships a bundled bugtrace-pentest runbook skill that codifies the scan workflow, focused modes, TOTP auth configuration, WAF bypass strategy, model shifting, and report formats. On top, a persona override enforces the operating rule: no scan without explicit user authorization, and every finding must be verified before it enters a report.

It is aimed at security researchers and blue-team members who already manage agents through dsh and need a repeatable pentest pipeline. If your current loop is run the CLI, paste results into a prompt, format the report by hand, this plugin turns that loop into a single start_scan call.