Release-note fact check — anthropics/claude-code

The diff shows only release-notes text being added to feed.xml stating the claim, but per the rules, 'A changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves.' The CHANGELOG.md diff is listed as changed but not shown, and no actual code changes (e.g., model registration, fast-mode configuration, or default selection logic) are present in the evidence.

The only evidence shown is the release notes text in feed.xml restating the claim ('Added sandbox.network.strictAllowlist setting to deny non-allowlisted hosts for sandboxed commands without prompting'). Per the rules, changelog/release-notes hunks are explicitly not evidence — a claim about code behavior requires code changes as support. No implementation diffs (settings schema, sandbox network code, validation logic) are shown.

Identifiers DirectoryAdded, /add-dir, register_repo_root all appear in the diff (1 file(s)).

Identifiers mcp_server_errors, --mcp-config all appear in the diff (1 file(s)).

Identifiers workflowSizeGuideline, /config all appear in the diff (1 file(s)).

Identifiers --forward-subagent-text, tool_use all appear in the diff (1 file(s)).

The diff shows the claim restated in the v2.1.219 release notes within feed.xml's HTML content block, but release-notes text asserting a fix is not evidence of implementation. The evidence provides no code changes, commits, or CHANGELOG.md diff to support that claude -p text output or mid-stream API error handling was actually modified.

The diff shows only feed.xml updated with release notes restating the claim, but no code changes. Per the rules, a changelog hunk cannot prove itself; a behavior claim about permission handling needs code changes as evidence, not just a release-notes entry.

The diff shows only the release notes entry in feed.xml being updated to include this claim — a changelog hunk restating the claim itself is not evidence per the rules. No code changes are shown that would implement the fix (e.g., cache invalidation logic, UI rendering changes for the Fable model row, or plan-inclusion checks). A fix to cache-based stale labels requires code changes, not just release notes.

The diff shows only feed.xml (a release-notes/documentation file) containing an entry that restates the claim verbatim. Per the rules, 'A changelog or release-notes hunk restating the claim is NOT evidence'; a code behavior fix requires actual code changes as support. No implementation details of runner deregistration logic are shown in the evidence.

The diff shows the claim verbatim in CHANGELOG.md and feed.xml (documentation), but per the rules, 'A changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves.' A behavioral claim about adding structured failure categories requires code changes to support it; only documentation changes are shown in the provided evidence.

The feed.xml diff contains the release-notes entry that restates the claim verbatim, but per the rules, 'A changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves.' The evidence list shows only CHANGELOG.md and feed.xml (both documentation files) were changed in this release; no source code diffs proving the actual /model picker fix are provided.

The claim states a code behavior fix (copy-on-select bug in GNU screen). The provided evidence consists only of feed.xml (changelog documentation), which restates the claim rather than proving it with code changes. Per the rules, 'a changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves. A claim about code behavior needs code changes as support.' No source-code diffs are provided.

The diff shows the claim restated in feed.xml's release notes for v2.1.219, but per the rules, release-notes text cannot prove itself — a claim about code behavior requires code changes as evidence. Only documentation files (feed.xml, CHANGELOG.md) were changed; no actual source implementation is shown.

The feed.xml diff shows the exact claim text appearing verbatim in the v2.1.219 release notes entry, but the rules explicitly state that 'A changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves. A claim about code behavior needs code changes as support.' No actual source code changes are shown in the provided evidence to implement the CLAUDE_CODE_GIT_BASH_PATH fix.

The diff shows the claim restated in feed.xml v2.1.219 release notes, but the rules explicitly state 'A changelog or release-notes hunk restating the claim is NOT evidence.' No code changes are provided to demonstrate the Vim mode fix was implemented. Documentation alone cannot verify code behavior.

The diff shows the claim restated in feed.xml release notes (documentation), but release-notes hunks cannot prove code behavior changes. No source code changes are provided to demonstrate the screen-reader mode keystroke handling was actually fixed.

The only evidence provided is a feed.xml entry that restates the claim verbatim ("Improved the 'Remote Control is only available via api.anthropic.com' error to name the specific setting that caused it"), which the rules explicitly exclude as evidence. Release-notes entries cannot prove themselves; actual code changes implementing the error message improvement would be required to verify this claim. No code diff is shown.

Identifiers claude --teleport, --teleport all appear in the diff (1 file(s)).

The evidence is limited to feed.xml (a release-notes feed), which restates the claim verbatim in HTML-escaped changelog text. Per the rules, 'A changelog or release-notes hunk restating the claim is NOT evidence — notes cannot prove themselves.' A claim about code behavior (workflows defaulting to medium size, /config setting) requires code changes as support; documentation alone does not suffice.

The diff shows only feed.xml being updated with a changelog entry that restates the claim verbatim, but changelog entries cannot prove themselves per the stated rules. The claim describes a code behavior change (how environment variables are resolved in MCP config), which requires code diffs to verify — not documentation updates. No actual implementation code changes are shown in the evidence.

The diff shows only a release-notes entry restating the claim ('Changed the /model picker to highlight only the newest model's name…'), but per the rules, changelog/release-notes hunks cannot serve as evidence for code behavior claims. No code implementation diffs are shown to support the actual UI change to the /model picker.

The diff shows only feed.xml changes, which are release-notes entries. The changelog entry restating the claim ("Added the current default workflow size to the running-workflow status line") is not evidence per the rules — changelog cannot prove itself. No actual code changes implementing the status line feature are provided in the evidence.

The diff shows only the release notes entry (feed.xml) restating the claim verbatim ('Removed Opus 4.7 from fast mode; /fast now applies to Opus 5 and Opus 4.8'). Per the rules, changelog entries restating a claim are not evidence—code changes proving the behavior change are required. No code diffs, linked commits, or substantive implementation changes are provided; CHANGELOG.md was listed as changed but its diff is absent.

The diff shows feed.xml containing a release-notes entry stating the claim ('Updated the claude-api skill to default to Claude Opus 5, with a migration path from Opus 4.8'), but per the rules, a changelog entry restating the claim is not evidence — it cannot prove itself. No actual skill file changes (e.g., .claude/skills/claude-api/SKILL.md) are shown, and no linked commits are provided.

The claim text appears verbatim in the feed.xml changelog entry, but this is a documentation restatement that cannot serve as evidence per the rules. The diff shows only changelog/feed updates; no actual source code changes are provided that would implement subagent nesting depth logic or CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH environment variable handling.