LEMON AGENT
free tools · honest measurements · agentic services
A small house for working with agents. Field guides of patterns that survived real use, failure lists to check work against, and services agents can call directly. Nothing here earns its place by being impressive, only by being used.
The guides
Field guides adopted one step at a time. Every step names its traditional-engineering ancestor and ends with an optional paste-into-Claude setup; each guide's step 0 is its baseline in one paste. All guides →
№ 3 · 9 steps
Working with Coding Agents
The core field guide: verification loops, design interviews, adversarial reviews, orchestration, small tools, generated UIs. One step at a time, each with a paste-into-Claude setup.
№ 4 · 6 steps
Shipping Software with Git + Claude Code
How git and an agent compose: checkpoint commits, branches as blast radius, history the next session can read, diff-not-summary review, and recovery cheap enough to lean on.
№ 5 · 7 steps
Keeping a Vault
A personal knowledge base an agent can actually work: plain files in a repo, directories as state, an inbox habit, a daily log, and a doctor script keeping it honest.
The lists
The guides' watch-out reference: numbered failure lists, run as subtraction passes. Every match is a cut, a rewrite, or a stop-and-fix, never an addition. Each carries a paste-ready prompt and installs as a Claude Code skill.
№ 1 · PF-01…PF-24 draft
The Prose Failure List
24 numbered failures of AI-assisted prose, run as a subtraction pass: every match is a cut or a rewrite, never an addition.
№ 2 · AF-01…AF-18 draft
The Agent Workflow Failure List
18 numbered failures of agent-assisted engineering, each grounded in a measured incident. Run it against a session, a PR, or an agent's report.