florianbruniaux/claude-code-plugins/check-cache-bugs

# Version check
claude --version
 
# All installed claude binaries
which -a claude 2>/dev/null
 
# Check if active binary is standalone or npm
file $(which claude) 2>/dev/null
ls -la $(which claude) 2>/dev/null

# Global config files
grep -r "cch=" \
  ~/.claude/CLAUDE.md \
  ~/.claude/MEMORY.md \
  ~/.claude/TONE.md \
  ~/.claude/FLAGS.md \
  ~/.claude/RULES.md \
  ~/.claude/RTK.md \
  ~/.claude/ANTI_AI.md \
  2>/dev/null
 
# Global skills, commands, agents, hooks (excluding this command file)
grep -rl "cch=" ~/.claude/skills/ 2>/dev/null
grep -rl "cch=" ~/.claude/commands/ --exclude="check-cache-bugs.md" 2>/dev/null
grep -rl "cch=" ~/.claude/agents/ 2>/dev/null
grep -rl "cch=" ~/.claude/hooks/ 2>/dev/null
 
# Project-level config
grep -r "cch=" CLAUDE.md .claude/CLAUDE.md .claude/MEMORY.md 2>/dev/null
grep -rl "cch=" .claude/skills/ 2>/dev/null
grep -rl "cch=" .claude/commands/ --exclude="check-cache-bugs.md" 2>/dev/null
grep -rl "cch=" .claude/agents/ 2>/dev/null
grep -rl "cch=" .claude/hooks/ 2>/dev/null
 
# Broader scan: all CLAUDE.md files across projects
find ~ -name "CLAUDE.md" \
  -not -path "*/node_modules/*" \
  -not -path "*/.git/*" \
  2>/dev/null | xargs grep -l "cch=" 2>/dev/null

# settings.json
grep -i -- "--resume\|--continue" ~/.claude/settings.json 2>/dev/null
grep -i -- "--resume\|--continue" .claude/settings.json 2>/dev/null
 
# Hooks
grep -rn -- "--resume\|--continue" ~/.claude/hooks/ 2>/dev/null
grep -rn -- "--resume\|--continue" .claude/hooks/ 2>/dev/null
 
# Commands and skills
grep -rn -- "--resume\|--continue" ~/.claude/commands/ 2>/dev/null
grep -rn -- "--resume\|--continue" .claude/commands/ 2>/dev/null
 
# Shell profiles (aliases, functions)
grep -n -- "--resume\|--continue" ~/.zshrc ~/.bashrc ~/.bash_profile ~/.zprofile 2>/dev/null
 
# Project scripts
find . -name "*.sh" -o -name "Makefile" 2>/dev/null | \
  xargs grep -l -- "--resume\|--continue" 2>/dev/null

# Global settings
grep -i "CLAUDE_CODE_ATTRIBUTION_HEADER\|ENABLE_TOOL_SEARCH" \
  ~/.claude/settings.json 2>/dev/null
 
# Project settings
grep -i "CLAUDE_CODE_ATTRIBUTION_HEADER\|ENABLE_TOOL_SEARCH" \
  .claude/settings.json 2>/dev/null
 
# Shell profiles
grep -i "CLAUDE_CODE_ATTRIBUTION_HEADER" \
  ~/.zshrc ~/.bashrc ~/.bash_profile ~/.zprofile 2>/dev/null

for b in $(which -a claude 2>/dev/null | sort -u); do
  echo "=== $b ==="
  $b --version 2>/dev/null || echo "unavailable"
  file $b 2>/dev/null
  ls -la $b 2>/dev/null
done

## Claude Code Cache Bug Audit: CC#40524
 
**Date**: [today]
**Active claude version**: [version]
**Install method**: [standalone binary | npm/npx | mixed]
 
---
 
### Bug 1: Sentinel replacement (standalone binary v2.1.36-v2.1.90)
**Status**: [FIXED (v2.1.91+) / SAFE / AT RISK / NOT APPLICABLE]
 
**Version >= v2.1.91**: [YES -> FIXED | NO -> check below]
**Mechanism active (standalone binary, v2.1.36-v2.1.90)**: [YES / NO]
**Trigger sentinel found in static config**: [YES (locations) | NO]
 
[If FIXED] Running v2.1.91+. Sentinel mechanism patched; npm and standalone are equivalent. No action needed.
[If AT RISK (v < 2.1.91 AND standalone AND sentinel found)]
-> Update to v2.1.91+: `npm install -g @anthropic-ai/claude-code@latest`
-> Temporary: remove `cch=00000` string from the flagged files, or use npm instead.
[If SAFE] Mechanism present in binary but no trigger in static config. Update to v2.1.91 still recommended.
[If NOT APPLICABLE] npm/npx install: Bug 1 mechanism absent.
 
---
 
### Bug 2: Cache prefix mismatch on --resume / --continue (v2.1.69-v2.1.89)
**Status**: [FIXED (v2.1.90+) / AT RISK (automated) / AT RISK (manual) / NOT APPLICABLE]
 
**Version >= v2.1.90**: [YES -> FIXED | NO -> check below]
**Automated usage found (hooks/settings)**: [YES (locations) | NO]
**Manual usage found (commands/skills/scripts)**: [YES (locations) | NO]
 
**Root cause**: JSONL writer stripped `deferred_tools_delta` records before writing to disk. On `--resume`,
the deferred tools layer had no announcement history and re-announced all tools from scratch, shifting every
message position and breaking messages-level cache prefix. Each resume rebuilt 87-118K tokens as cache_creation.
Anthropic tracked as inc-4747. Fixed in v2.1.90.
 
[If FIXED] Running v2.1.90+. Bug resolved. `--resume` usage found in config is now safe at the cache level.
[If AT RISK]
-> Update to v2.1.90+: `npm install -g @anthropic-ai/claude-code@latest`
-> Immediate: avoid `--resume` and `--continue` until updated.
[If NOT APPLICABLE] Version < 2.1.69: not affected.
 
---
 
### Bug 3: Attribution header per-session hash (v2.1.69+, no fix yet)
**Status**: [ACTIVE / MITIGATED / NOT APPLICABLE]
 
**Version in affected range (>= v2.1.69)**: [YES / NO]
**CLAUDE_CODE_ATTRIBUTION_HEADER set to false/0/no/off**: [YES -> MITIGATED | NO -> ACTIVE]
 
[If ACTIVE] Every session start and every subagent call misses the system prompt cache (~12K tokens
rebuilt at cache_creation rate). A session with 5 subagents = 6 cold misses.
-> Fix (immediate, no restart needed): add to ~/.claude/settings.json:
  {
    "env": {
      "CLAUDE_CODE_ATTRIBUTION_HEADER": "false"
    }
  }
-> Expected impact: cache hit ratio 48% -> ~99.98% (measured, source: @whiletrue0x CC#40524)
 
[If MITIGATED] Header already disabled. No action needed.
[If NOT APPLICABLE] Version < 2.1.69: not affected.
 
---
 
### B4/B5: Context mutation (informational, all versions)
 
These bugs are controlled by server-side GrowthBook flags and cannot be mitigated client-side.
No config check is possible; report as informational.
 
**B4: Microcompact**: Three mechanisms silently replace tool results with `[Old tool result content cleared]`
before sending to the API. The `/export` command shows original context, NOT what the model received.
Workaround: start fresh sessions periodically to reset the tool result pool.
 
**B5: Tool result budget cap (200K chars aggregate)**: Built-in tools (Read, Bash, Grep, Glob, Edit)
hit a 200K character cap on aggregate tool results per session. Results are truncated beyond this threshold.
MCP tools can partially override via `_meta["anthropic/maxResultSizeChars"]` (v2.1.91+, up to 500K).
 
---
 
### Multiple binaries
[List each binary found, version, type (standalone/npm), and per-bug status]
 
---
 
### Summary