Home / Guides
Guides
Practical write-ups of the patterns behind each pack — the hook contract, the code, the failure modes — with the exact snippets you would ship.
-
2026-10-11 · 752 words · Bitter Pill Audit
Auditing your CLAUDE.md: cut, keep, sharpen
A repeatable audit for instruction files — flag vague, redundant, default and contradictory rules, estimate token cost, rewrite the survivors shorter.
-
2026-10-11 · 764 words · Credential Guard
How to block credential leaks in Claude Code with a PreToolUse hook
A PreToolUse hook that denies curl, wget and HTTP tool calls carrying a literal API key, lets $VAR references through, and logs every decision locally.
-
2026-10-11 · 791 words · Reproduce First
Reproduce before you fix: a bug-report workflow for coding agents
Make a captured failure — command output, failing test or screenshot — a hard prerequisite before any edit, and reuse the same capture to prove the fix.
-
2026-10-11 · 722 words · Research Fleet
Running a research fleet with Claude Code subagents and verifying every URL
Split a question into sub-questions, run parallel subagents, cross-check claims into confidence tags, and fetch every link before it reaches the report.
-
2026-10-11 · 745 words · Release Scrub
Scrubbing a repo for personal data before you publish
Scan a folder for emails, phones, checksum-valid IBANs, API keys and a private names list, triage every finding, and publish only when a re-run is clean.
-
2026-10-11 · 744 words · Evidence Gate
Stop Claude Code from marking work done without evidence (the Stop hook pattern)
Use a Stop hook to refuse completion while checklist items are unchecked or lack an evidence line, with the exact missing criteria shown to Claude.