A collection of 61 posts
Use /cost as an engineering checkpoint in Claude Code: notice when a session is sprawling, then narrow the next proof, switch strategy, or stop before more tokens become more confusion.
Read more
Use Claude Code to keep a decision log during real work: record assumptions, rejected options, evidence, and corrections so the final diff is easier to review and safer to ship.
Read more
Debug auth and deployment failures with Claude Code without pasting secrets: map variable names, redact logs, inspect config paths, and make the final diff prove no credentials leaked.
Read more
Use Claude Code as a migration safety partner: inspect the current schema first, demand a boring migration plan, run dry checks, verify app code, and write the rollback note before deployment.
Read more
A field guide to making Claude Code declare the files it will not touch before editing, then verifying the diff against that boundary after the first patch.
Read more
Use /status before a Claude Code session turns into autopilot: check the session state, name what has been proven, and decide whether to continue, compact, clear, or stop.
Read more
Use /model as a deliberate workflow choice: match model strength to task risk, evidence needs, and verification instead of always choosing the biggest option.
Read more
Use /memory for durable team rules—verification commands, style constraints, deployment cautions—not for every theory from today’s debugging session.
Read more
Use /compact as a deliberate checkpoint in long Claude Code sessions: keep the bug, evidence, decisions, diff, and verification path while dropping the noise.
Read more
When a check fails, do not ask Claude Code to “fix it.” Paste the exact failure, freeze the scope, request one hypothesis, and make the next patch prove that signal changed.
Read more