agent-parser

Parse pentest tool output into a SQLite world model and Markdown vault.

Updated May 27, 2026
One-click install
npx skills add https://github.com/baljinnyamday/autonomous-red-team-agent --skill agent-parser-baljinnyamday
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-parser
Source: https://github.com/baljinnyamday/autonomous-red-team-agent/tree/main/agent-baseline/.agents/skills/agent-parser
Command: npx skills add https://github.com/baljinnyamday/autonomous-red-team-agent --skill agent-parser-baljinnyamday

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Raw output from pentest tools like nmap, hydra, and impacket is unstructured text that an agent cannot reason over. This Skill converts that output into typed, structured facts stored in a SQLite index and re-exported as a Markdown vault, so planning always happens against current, normalized data. ## Core Features & Use Cases - Multi-tool parsing: Ingests output from nmap (XML), ssh_exec, hydra, impacket (psexec, secretsdump, wmiexec, smbexec), and msfrpc into hosts, services, credentials, sessions, and observations. - Stdin streaming: Pipes tool output directly into the parser (tool | agent-baseline parse <tool> -) to avoid stale intermediate files and race conditions. - Vault re-export: Regenerates the Markdown vault from SQLite via export-obsidian after every parse, keeping the planner-readable surface in sync. - Use Case: After running nmap -oX - 10.0.0.0/24, stream the XML into the parser, re-export the vault, and immediately plan the next attack step from up-to-date host and service facts. ## Quick Start Pipe the nmap XML output you just captured into the agent-baseline parser with a fresh SQLite database, then re-export the Markdown vault and report the parse_status.

Frequently Asked Questions about agent-parser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I parse nmap output into a structured format?

Run nmap with XML output (`nmap -oX - <target>`) and pipe it into `agent-baseline --db <path> parse nmap -`. The parser extracts hosts and services with port, protocol, product, and version into SQLite. Plain text nmap output is not supported; only the `-oX` XML format parses.

What pentest tools can the agent-baseline parser ingest?

The parser supports nmap (XML), ssh_exec, hydra, impacket tools (psexec, secretsdump, wmiexec, smbexec), and msfrpc. Unsupported tools return a fallback result: no facts are extracted, but the raw output is still archived and the action is logged.

Why is my Markdown vault showing stale data after parsing?

Parsing writes only to SQLite; the vault is not updated automatically. You must run `agent-baseline --db <path> export-obsidian --out <vault>` after every parse. Treat parse and export as a single unit so planners never read outdated facts.

Can I re-ingest the same tool output twice?

Yes, ingestion is idempotent at the SQLite layer. Duplicate facts collapse on host, service, and credential identity, so retrying a parse is safe. Re-export the vault afterwards to reflect the final state.

Should I query SQLite directly for planning instead of the vault?

No. The `snapshot` and `search` commands read SQLite directly but are intended only as debugging aids when the vault export is missing expected data. All planning reads should go through the exported Markdown vault, treating SQLite as opaque infrastructure.