Skip to content
DevAgentRadarDevAgentRadar
Sign in
Proposals/Drop legacy managed MCP configs and shadowed pla
proposaleveryoneP5High impactxAI Grok CLI (Grok Build)

Drop legacy managed MCP configs and shadowed platform skills in Grok CLI

Grok CLI 8a14c91 removes the legacy managed MCP configs client from shell/pager and deletes stale extracted platform skills that shadow bundled ones, plus faster /resume, worktree flag keep, and skill-path hints on failed reads.

Why this loop

This monorepo sync is tagged breaking and explicitly removes the legacy managed MCP configs client from shell and pager, so lingering managed MCP setup will stop working. It also removes stale extracted platform skills that shadow bundled skills—leaving old extractions in place can silently override the bundled skill set. Adjacent fixes (registered skill path suggestions on failed reads, standalone worktree flag kept across cwd switch/resume, faster local /resume on large transcripts, bounded .envrc evaluation) only pay off if you clear the removed paths and exercise the new resume/worktree behavior.

Proposed actions

  1. Search the repo and user config for legacy managed MCP config references used by Grok shell/pager and delete or rewrite them to non-legacy MCP wiring after pulling 8a14c91
  2. Delete stale extracted platform skill directories that shadow bundled skills so Grok loads the bundled skill set instead of old extractions
  3. From a large existing session, run local /resume and confirm transcript reload completes on the sped-up path
  4. Switch cwd inside a standalone grok worktree, resume the session, and confirm the worktree flag and branch display still show the standalone grok worktree
  5. On a failed skill read, retry using the registered skill path the CLI now suggests in the error output

Agent prompt

Paste into your agent or query via MCP (`get_agent_prompt`) — free, no extra AI cost

Paste into Claude Code / CLAUDE.md task

# DevAgentRadar → Claude Code

You are helping me adopt a real coding-assistant change. Work only from the facts below. Do not invent features.

## Context
Assistant: xAI Grok CLI (Grok Build)
Proposal: Drop legacy managed MCP configs and shadowed platform skills in Grok CLI
Summary: Grok CLI 8a14c91 removes the legacy managed MCP configs client from shell/pager and deletes stale extracted platform skills that shadow bundled ones, plus faster /resume, worktree flag keep, and skill-path hints on failed reads.
Primary source: https://github.com/xai-org/grok-build/commit/8a14c91d88875a831a38b3a066b1683116bcb31c

## Why it matters
This monorepo sync is tagged breaking and explicitly removes the legacy managed MCP configs client from shell and pager, so lingering managed MCP setup will stop working. It also removes stale extracted platform skills that shadow bundled skills—leaving old extractions in place can silently override the bundled skill set. Adjacent fixes (registered skill path suggestions on failed reads, standalone worktree flag kept across cwd switch/resume, faster local /resume on large transcripts, bounded .envrc evaluation) only pay off if you clear the removed paths and exercise the new resume/worktree behavior.

## Suggested actions
1. Search the repo and user config for legacy managed MCP config references used by Grok shell/pager and delete or rewrite them to non-legacy MCP wiring after pulling 8a14c91
2. Delete stale extracted platform skill directories that shadow bundled skills so Grok loads the bundled skill set instead of old extractions
3. From a large existing session, run local /resume and confirm transcript reload completes on the sped-up path
4. Switch cwd inside a standalone grok worktree, resume the session, and confirm the worktree flag and branch display still show the standalone grok worktree
5. On a failed skill read, retry using the registered skill path the CLI now suggests in the error output


## Your job
1. Restate the change in one sentence.
2. Propose a minimal plan for my repo (or a throwaway pilot).
3. Implement only what I approve; prefer small diffs and tests.
4. Call out risks (permissions, breaking APIs, cost).

Start by confirming you understood the proposal.
mcpmodelclibreakingRelease source ↗

Your loop

This browser · no sign-in · not shared as “you”

Community signal · 0 approved · 0 deferred · 0 ignored

Anonymous public tally (one vote per visitor; changing choice updates your vote). Not moderated product truth.

Originating release signal

xAI Grok CLI (Grok Build)8a14c91Aug 8, 2026

Synced from monorepo

Synced from monorepo · Changes: · Make memory trace wait signal-safe · Keep standalone worktree flag across cwd switch and resume · Detect standalone grok worktrees for branch display · Allow Send Now throughout goal mode · Make non-blocking startup structural · A requested quit always exits the process · Suggest registered skill paths on failed reads · Diagnose scroll anchor jolts · Answer HITL ExtMethods on -p · Bound .envrc evaluation so a blocked read can't freeze session load · Protect the tools server from the OOM killer and attribute OOM kills · +5 more changes

Verified excerpt — the source's own words
Synced from monorepo

Changes:
- Make memory trace wait signal-safe
- Keep standalone worktree flag across cwd switch and resume
- Detect standalone grok worktrees for branch display
- Allow Send Now throughout goal mode
- Make non-blocking startup structural
- A requested quit always exits the process
- Suggest registered skill paths on failed reads
- Diagnose scroll anchor jolts
- Answer HITL ExtMethods on -p
- Bound .envrc evaluation so a blocked read can't freeze session load
- Protect the tools server from the OOM killer and attribute OOM kills
- Speed up local /resume on large session transcripts
- Drain subagents before session delete
- Remove stale extracted platform skills that shadow bundled skills
- Honest system.notify acks for synchronous pre-forward drops
- Remove legacy managed MCP configs client from shell and pager
Primary source ↗