Prompt, Skill, Memory, Agent
Prompt, skill, memory, tool, and agent are five layers with different lifespans. Here is the test for which one a fact belongs in.
Read guide →Monolith field manual / Open access
Build agents you can explain, control, and verify.
Plain-language guides for moving from ad hoc prompting to repeatable agent work. Start with the operating model, then add tools only where the job requires them.
All guides / 12 operating notes
From first setup to controlled execution.
Each guide answers one practical decision, with the tables, commands, and checklists to act on it. No prompt-count theater, no universal compatibility claims, and no requirement to hand an agent unnecessary permissions.
Install SKILL.md Safely
A ten minute review for any SKILL.md: stage the package, read the frontmatter, grep the scripts, install to the right host path, then rehearse it.
Read guide →Beginner Setup Checklist
A first AI agent setup in order: pick one host, bound the workspace, write deny rules before allow rules, cap spend, gate consequential actions.
Read guide →Hermes, OpenClaw, Claude Code
Hermes, OpenClaw, and Claude Code compared on install, skill format, memory, approvals, cost, and platforms, plus which one to run.
Read guide →Human Approval Gates
A risk ladder, the verbs to gate, the harness settings that enforce them, and the four-case drill that proves a gate fires.
Read guide →Turn an SOP Into a Skill
How to convert a written SOP into an agent skill: evidence gathering, decision rules, folder layout, routing metadata, approval gates, and tests.
Read guide →Agent Workflows for Freelancers
Five agent workflows a freelancer can run this week: meeting follow-up, proposals, delivery QA, research, and the weekly review, all behind human gates.
Read guide →Agent Workflows for Small Business
Five AI agent workflows a small business can run this month, with the approval gate on each, the evidence to check, and the failures to expect.
Read guide →Audit Skills for Prompt Injection
A working audit for prompt injection in agent skills: what to grep, which channels to map, which actions to gate, and the fixtures to rerun.
Read guide →Control Agent Costs
Cut AI agent API costs with prompt caching, input hygiene, task-based model routing, bounded loops, and a cost-per-finished-task review.
Read guide →Choose Tools and Permissions
Pick the smallest AI agent toolset, sort every action into read, draft, or execute, and keep sends, spends, and deploys behind a rule.
Read guide →- Guides
- 12one decision each
- Reading
- 74 min14,726 words at 200 wpm
- Updated
- Sep 3, 2026facts checked on edit
- Cost
- Freeopen access, no sign-up
Free foundation
Turn the reading into a safe first setup.
Monolith Agent Ready converts the principles into a setup brief, safety gates, a cost-and-tools plan, and three first workflow specifications.
Get Agent Ready Free