What problem does it solve?
During long penetration testing engagements, context compression and multi-agent coordination cause critical findings, credentials, and attack-chain details to be lost. This Skill enforces a disciplined write-as-you-go workflow that persists every confirmed fact, vulnerability, and relationship edge into a project-bound SQLite blackboard instead of relying on conversation memory.
Core Features & Use Cases
- Fact Graph Management: Upsert project facts with structured fact_key categories (target, auth, infra, finding, chain, exploit, poc), confidence levels (confirmed, tentative, deprecated), and relationship links forming a queryable attack graph.
- Vulnerability Recording: Register deliverable findings via record_vulnerability with title, severity, type, target, POC, impact, and remediation, with deduplication through list_vulnerabilities.
- Multi-Agent Coordination: Coordinators aggregate 'pending write' items from sub-agents without tool access and persist them, ensuring no evidence is lost across sessions.
- Use Case: During an authorized web app pentest, after confirming a SQL injection on the login endpoint, immediately upsert a finding/sqli-login fact with full reproduction steps and links to the target fact, then record the formal vulnerability before continuing exploitation.
Quick Start
Ask the agent to record the currently confirmed finding as a project fact with confidence level and relationship links, then register it as a formal vulnerability in the bound project blackboard.