theorydeltaclaim-fidelity audits
built 2026-07-17dossiers: 5last verified 2026-07-09independent · evidence-traced · no vendor influence
← back to Claude Code dossier

CLAIM 4 / Claude Code

"Hooks are user-defined shell commands, HTTP endpoints, or LLM prompts that execute automatically at specific points in Claude Code's lifecycle"

partial

Observed

Hooks fire in the common path, but 30+ failure modes are confirmed across silent non-firing, ignored decisions, platform breakage, and data corruption: PreToolUse/PostToolUse hooks intermittently never fire while other events in the same settings file fire reliably (#6305, open, has repro, root cause unknown); all plugin-registered hooks stop executing after context compaction (#25655, closed not-planned); SessionStart HTTP hooks never dispatched in v2.1.63 (#30170); and PostToolUse hooks were observed not firing on Write-tool invocations. No single hook event is a reliable enforcement point — defense-in-depth across multiple events is required.

  1. UPDATE 2026-06-28: the hook failure surface grew by 4 issues in 17 days — non-interactive non-firing (#71022), Linux transcript gap (#70632), /clear session_id corruption (#70606), and unreliable hookAdditionalContext (#70479).

Evidence

Verification

verified 2026-07-16 · by probe-ci · probe probes/claude-code/claim-4/

theorydelta.com · 2026independent · evidence-backed · every claim sourced or labelledabout ·glossary ·rss ·mcp ·llms.txt