agent-memory-vault

Query Markdown vault exports of hosts, credentials, and actions from agent engagement memory.

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

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Agents working over a red-team engagement memory store can waste context loading an entire SQLite world model or full vault when they only need one host, credential, or action. This Skill provides a targeted query workflow over the folder-based Markdown/Obsidian vault produced by agent-baseline export-obsidian, so agents read only the files they need. ## Core Features & Use Cases - Targeted search-first workflow: Use rg, fd, and find to locate hosts, services, credentials, and action records by IP, hostname, port, username, or action id before opening any file. - Structured vault layout: Per-host folders expose index.md, services.md, creds.md, actions.md, reached-by.md, and raw.md, plus cross-host credential and chronological action indexes. - Pivot and replay reasoning: Read reached-by.md before pivoting and actions.md to avoid repeating failed attempts; open raw.md only when summaries are ambiguous. - Use Case: An agent needs to know which credentials are valid against host 10.0.0.5 before pivoting. It runs rg "Status: valid" over the vault, reads that host's creds.md and reached-by.md, and plans the next move without loading the full database. ## Quick Start Ask the agent to search the exported vault for a specific host IP and summarize its services, valid credentials, and how it was reached.

Frequently Asked Questions about agent-memory-vault

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

FAQPage Schema
How do I query an Obsidian-style Markdown vault for a specific host?

Search for the host's IP or hostname with rg across the vault first, then open only that host's folder. Read index.md, services.md, and creds.md for the overview, and reached-by.md before planning any pivot.

How do I find valid credentials across multiple hosts in a vault?

Run rg for the valid status marker over the credentials directory and host folders, for example rg "Status: `valid`" on the vault. The credentials/index.md file also provides a cross-host credential inventory.

Can I edit vault Markdown files directly to update agent memory?

No. The vault is an export, not the source of truth. Ingest new tool output into the SQLite database with the parse command, then re-run export-obsidian so the Markdown reflects parsed facts.

When should I read raw.md files in the vault?

Read raw.md only when the summary files are ambiguous, incomplete, or need evidence. Raw files contain large command output snippets and should not be loaded by default to conserve context.

What if fd is not available for searching the vault?

Fall back to find for path discovery, as shown in the workflow. rg remains the primary content search tool, and find can substitute for fd when locating files like services.md across host folders.