Author
Ren builds agent infrastructure and writes copy-paste tutorials for engineers shipping LLM tool-use systems.
A grep bisect tells you when today's code landed, not when the behaviour changed. Here is the argv-capture harness that dates it properly, and the worked example where grepping made a maintainer look wrong and running the versions proved him right.
Since @anthropic-ai/claude-agent-sdk 0.3.257 a malformed hook entry is deleted from settings.json and the rest of the file loads normally, so the hook just never runs. PreToolUse and PermissionRequest are the two exceptions, where the whole file is voided. Bisected against 0.3.252, with a runnable detector.
TypeError: Invalid http_client argument means the anthropic Python SDK moved to httpx2 in 1.0.0. The rename is one import. The parts nobody mentions are that the guard printing that message only exists from 1.4.0, and that it cannot see your respx or OpenTelemetry layer at all (2026).
Set _meta["anthropic/requiresUserInteraction"] to true and Claude Code prompts for that tool on every call, even in bypassPermissions, and even when an allow rule matches. It shipped in v2.1.199 with no changelog entry, which is why almost nobody has written about it.
Claude Code v2.1.251 stopped project settings setting CLAUDE_CODE_TMPDIR, CLAUDE_CONFIG_DIR, HOME, TMPDIR and the XDG family from a .claude/settings.json env block. Nothing errors. Here is the full ignore list, the four scopes it splits into, and the debug flag that shows you the drop.
permissions.blockReadsOutsideWorkingDirectories blocks reads outside your working directories in every permission mode, including bypassPermissions. Writing false does not turn it off, because the key is a logical OR across all settings sources. Plus the 53-hour macOS defect window, dated from the npm registry.
"Checkpointer requires one or more of the following 'configurable' keys" means you compiled with a checkpointer and invoked without one. The message names three keys and only thread_id works. Measured in Python and TypeScript (2026).
MCP elicitation now has two modes, and the specification revision Google ranks first is superseded. What form mode allows, why URL mode is mandatory for credentials, why ctx.elicit() stopped working on 2026-07-28, and a runnable schema linter in Python and TypeScript.
Anthropic merged custom slash commands into skills in 2026. Your .claude/commands files still work, nothing is deprecated, and the frontmatter you gain is the reason to move. Includes the command-name rule that catches everyone and a runnable collision audit.
Four launch-week tutorials carried a stored publication timestamp that predated the registration of agentnotebook.dev. The timestamps are corrected, the tutorial text is unchanged, and the check is one you can run yourself.
Claude Code has two memory systems and only one of them is capped. A precise 2026 walkthrough of CLAUDE.md load order, the auto memory directory, the 200-line-or-25KB index limit, and a runnable linter that measures your MEMORY.md the way Claude Code measures it.
Plan mode is not a read-only state. It is a rule at step 4 of a six-step permission evaluation, which is why allow rules stop applying while you plan and why a session with bypass permissions can edit anyway. The CLI keystrokes, the settings, the Agent SDK equivalent, and a plan-then-execute pipeline in Python and TypeScript. August 2026.
