The agent hit an API error halfway through a job and nobody knew until morning. This setup records why it stopped, retries the errors that time fixes and stops on the ones that don't. About fifteen minutes to set up.
11 min read
Fifteen minutes to a two-line status bar showing unreviewed lines, the prompt-cache clock and how much context is left, so you stop the agent before the diff gets out of hand.
12 min read
A recorder, a generated test file and a mutation run give you a safety net you can measure, not just feel good about. About 20 minutes to set up.
12 min read
A 70-line script, a pre-commit hook and one CI step catch the frontmatter mistakes Claude Code skips without telling you. About 15 minutes to set up.
11 min read
One 50-line script stops AWS keys, GitHub tokens, private keys and connection strings leaving your terminal, and puts the repo state into every other prompt. About 15 minutes, and the gaps are listed.
12 min read
A redacting gather script, a read-only claude -p call with a JSON schema, and a checklist renderer. About 15 minutes to build, and the model is told to separate evidence from guesses.
12 min read
A /spec skill, a 50-line checker and one PreToolUse hook that stops the agent writing code outside a spec you've signed off. About 15 minutes to set up.
14 min read
