Follow Discord
Sweep 03 Nov 2026 · 22:27Z Build v2.1.318 536 read Stable v2.1.311 Latest v2.1.318 Next v2.1.318 Feeds RSS JSON llms.txt llms-full.txt Unofficial
Prompt changes · v2.1.318 against v2.1.317

6 changes since v2.1.317, landing in 5 of 21 prompts.

Every prompt v2.1.318 sent, compared word for word with the capture before it. Each card is one edit, with every prompt it reached; the table after them shows where each one landed.

What changed 6

1

System prompt · under “Context management”

EndConversation deferred-tool guidance removed

The system prompt no longer mentions the deferred EndConversation tool. It previously said to use it only for sustained user abuse aimed at the assistant or when the user explicitly asked for a demonstration, and to load its full guidance via ToolSearch first. This appeared in only 1 of 23 captures, so it likely applies to one specific setup.

Behind a feature flag

+0 −34 words

EndConversation (deferred tool): use only for sustained user abuse directed at the assistant, or when the user explicitly asks to see it demonstrated. Load the full guidance via ToolSearch("select:EndConversation") before using it.

2

Slash command text · under “Get Started”

Guide must note sessions left out of usage stats

A new instruction for the guide-creating slash command says that if `sessionsLeftOut` is above 0, Claude should state in one plain sentence under "How We Use Claude" that some sessions could not be tied to the creator's current sign-in, so the numbers may be incomplete. Claude must repeat that note when presenting the draft. It appeared in only 1 of 21 captures, so it likely applies to the setup that runs this command.

+55 −0 words

If `sessionsLeftOut` is above 0, that many sessions were left out because they couldn't be tied to the way the guide creator is signed in now, so the numbers may not cover everything from this period: say so in one plain sentence under "How We Use Claude" and again when you present the draft.

3

Slash command text · under “Usage data (last 30 days)”

Usage data gains a sessionProvenance field

The usage data block for the last 30 days now includes a `sessionProvenance` field whose value is a placeholder marking artifact content authored by others. It appeared in only 1 of 21 captures, tied to the slash command that receives this usage data.

+6 −0 words

"sessionProvenance": "<artifact-content-authored-by-others/>",

4

Slash command text · under “Usage data (last 30 days)”

Usage data gains a sessionsLeftOut count

The usage data block for the last 30 days now includes a `sessionsLeftOut` count, set to 0 in this capture. This is the field the new instruction checks before telling Claude to warn that the numbers may be incomplete. It appeared in only 1 of 21 captures.

+2 −0 words

"sessionsLeftOut": 0,

5

PushNotification tool parameters · line 16

PushNotification no longer requires a status parameter

The list of required PushNotification parameters drops "status" and keeps only "message", so Claude can send a notification without supplying a status. The diff does not show whether status remains available as an optional field. This landed in 3 of 6 captures, so it may apply only to some setups.

Behind a feature flag

+0 −1 words

"message",

"status"

6

Appended context (skills, environment) · at “EndConversation”

EndConversation dropped from appended tool listing

The name EndConversation was removed from the appended context, so it is no longer listed there as an available deferred tool. This was seen in only 1 of 23 captures and matches the removal of the EndConversation guidance from the system prompt.

Behind a feature flag

+0 −1 words

EndConversation

Where each change landed

Every prompt the release sent, one row each. A number is a change from the list above; click it to jump there. Rows with more than one model string count them, because each model gets its own copy of the prompt.

Prompt 1 2 3 4 5 6 Status
Main prompt
Claude Code, default27 models · · · · · · unchanged
Claude Code, seeded with this account's feature flags27 models 13/27 · · · 4/27 13/27 3 changes
Claude Code, seeded with anonymous feature flags27 models · · · · 4/27 · 1 change
Agent SDK, default27 models · · · · · · unchanged
Agent SDK, seeded with this account's feature flags27 models · · · · 4/27 · 1 change
Agent SDK, seeded with anonymous feature flags27 models · · · · 4/27 · 1 change
Auto-mode classifier
Auto-mode classifier1 model · · · · · · unchanged
Slash commands
/init1 model · · · · · · unchanged
/security-review1 model · · · · · · unchanged
/team-onboarding1 model · ● ● ● · · 3 changes
Compaction
Compaction instruction1 model · · · · · · unchanged
System reminders
Claude md1 model · · · · · · unchanged
Hook pretool block1 model · · · · · · unchanged
Read empty1 model · · · · · · unchanged
Read offset1 model · · · · · · unchanged
Skill run1 model · · · · · · unchanged
Todo nag1 model · · · · · · unchanged
Toolsearch deferred1 model · · · · · · unchanged
Subagents
explore1 model · · · · · · unchanged
general-purpose1 model · · · · · · unchanged
plan1 model · · · · · · unchanged

Did not change

30 tool descriptions are word for word the same, 1 moved and is listed above.

Agent AskUserQuestion Bash CronCreate CronDelete CronList DeferredToolPlaceholder Edit EnterPlanMode EnterWorktree ExitPlanMode ExitWorktree ListAgents Monitor NotebookEdit Read ReportFindings ScheduleWakeup SendMessage Skill TaskCreate TaskGet TaskList TaskStop TaskUpdate ToolSearch WebFetch WebSearch Workflow Write

16 prompts did not move at all: Claude Code, default, Agent SDK, default, Auto-mode classifier, Slash commands: /init, Slash commands: /security-review, Compaction: Compaction instruction, System reminders: Claude md, System reminders: Hook pretool block, System reminders: Read empty, System reminders: Read offset, System reminders: Skill run, System reminders: Todo nag, System reminders: Toolsearch deferred, Subagents: explore, Subagents: general-purpose, Subagents: plan.

Set aside as capture noise

These change on every capture by construction, so they are masked before comparing and never counted as a change.

  • 354 billing header lines
  • 220 capture date lines
  • 220 host kernel lines
Feedback