Guides
Guides for long-running coding agents
Practical playbooks for long-running, unattended coding: overnight runs, quality, coverage, defects, dependencies, security, accessibility, and documentation.
Start with your situation
Run Claude Code, Codex or Cursor overnight
Leave a coding agent working overnight safely: a bounded task list, the right permissions, an awake machine, crash recovery, and a morning review.
Read page →02 · GuidesMake Claude Code, Codex or Cursor finish tasks
Keep your coding agent working until every approved task is done: outcomes with definitions of done, a stop hook that returns open work, and receipts.
Read page →03 · GuidesImprove your product with a coding agent
Let Claude Code, Codex, or Cursor research your product, rank opportunities, and ship the strongest complete improvements within a fixed clock.
Read page →04 · GuidesAutomatically find the best work in a project
Give Automatic Hunt a sentence and a time budget; it inspects the project and composes the catalog work that serves your objective without inventing a backlog.
Read page →05 · GuidesRun a cited SEO audit with a coding agent
Audit technical SEO, search intent, structured data, crawlability, internal links, accessibility, and performance under a finite evidence contract.
Read page →06 · GuidesResearch and synthesize a list of sources
Give a coding agent approved URLs or local sources and get a cited, conflict-aware report with facts separated from inference.
Read page →07 · GuidesWrite source-grounded documentation
Use a coding agent to turn authoritative references into clear documentation with cited claims, verified links, examples, and visible uncertainty.
Read page →08 · GuidesUse a coding agent to find and fix quality debt
Survey tests, code, accessibility, contracts, docs, dependencies, and security, then work only applicable findings.
Read page →09 · GuidesUse a coding agent to find and fix bugs
Use a coding agent to find and fix bugs: review, fix, verify, and re-review defects until a clean pass or the shift deadline.
Read page →10 · GuidesRaise test coverage with a coding agent
Let Claude Code, Codex, or Cursor raise test coverage overnight with meaningful tests around risk and behavior, using coverage as a tripwire, not a target.
Read page →11 · GuidesAutomate dependency upgrades safely
Upgrade applicable dependencies with compatibility checks, focused commits, and a bounded rollback path.
Read page →12 · GuidesRun a bounded coding-agent security review
Inspect applicable security risks, verify fixes, preserve confidential details, and leave reviewable local evidence.
Read page →13 · GuidesFix accessibility issues with a coding agent
Find and fix applicable keyboard, semantic, contrast, labeling, motion, and responsive accessibility problems.
Read page →14 · GuidesRepair documentation drift
Use a coding agent to align public docs, examples, commands, and configuration references with the product that actually ships.
Read page →15 · GuidesTurn selected GitHub issues into a shift
Import explicitly chosen issues as quoted source, review them, and promote only approved work into the punch list.
Read page →16 · GuidesKeep Codex on task after context compaction
Give Codex an on-disk continuation point before a long session compacts: completed work, rejected paths, exact next action, and verification still due.
Read page →17 · GuidesClaude Code stops early: what to check
Distinguish an early Claude Code completion from permission waits, API failures, owner stops, and limits before changing a Nightshift shift.
Read page →18 · GuidesCodex stops before finishing: what to check
Diagnose a Codex task that ends with work still open: check the active contract, permissions, rollout, SessionEnd, and genuine service or usage limits.
Read page →19 · GuidesCursor Agent stops early: what to check
Diagnose a Cursor Agent that ends with open work: distinguish normal completion, Stop-button abort, question wait, error, and IDE-to-CLI recovery.
Read page →20 · GuidesClaude Code API Error 500: diagnose and recover
Check whether a Claude Code 500 or 529 came from the service, whether the shift is still armed, and what the Nightshift watchman actually resumed.
Read page →21 · GuidesPark questions instead of waiting overnight
Claude Code, Codex, or Cursor stopped at 02:40 with a question and waited until morning. Nightshift denies the question tool on shift and parks a default.
Read page →22 · GuidesRepair API contract drift
Use a coding agent to align checked-in server, schema, client, fixture, and generated API artifacts without silently changing the public contract.
Read page →23 · GuidesClean CI warnings without hiding them
Use a coding agent to fix repository-owned compiler, test, bundler, and deprecation warnings at their cause while reporting external warnings honestly.
Read page →24 · GuidesRemove dead code only with evidence
Let a coding agent use the repository's own analyzer to remove unused exports, files, branches, or dependencies without deleting dynamic or external paths.
Read page →25 · GuidesRepair flaky tests without weakening them
Use a coding agent to reproduce intermittent failures within a declared budget, fix demonstrated causes, and report unreproduced suspects honestly.
Read page →26 · GuidesRepair localization key drift
Use a coding agent to align missing, unused, duplicate, placeholder, plural, and structural locale keys without inventing translations.
Read page →27 · GuidesResolve TODO and FIXME debt responsibly
Turn actionable code comments into verified work while staging ambiguous product and compatibility decisions for the owner.
Read page →