Skip to content
DevAgentRadarDevAgentRadar
Sign in
Guidance/Daily Briefing
digesteveryonedigestevidenceradarmodelagent7 min read · Aug 7, 2026

Radar brief — 2026-08-07

Agent plugins, self-hosted runners, and plan-mode locks land together — while Goose ships a real RCE advisory and Codex tightens cyber-model approvals

What the evidence shows

High-signal, versioned releases from the window (GitHub/changelog-weighted; 55% trusted mix):

Claude Codev2.1.224
Key Signal & Impact
Self-hosted runner turns your machines/containers into session backends (Team/Enterprise); archive plugin install over HTTPS with optional SHA-256 pin; Bedrock region-prefix env; broad agent/MCP/permission surface churn
OpenAI Codex CLIrust-v0.147.0
Key Signal & Impact
Portable Agent Plugins + multi-catalog search; conversation sections; --approve-for-me auto-reviewed approvals; import/sync Cursor & Claude skills without dupes
OpenAI Codex CLIrust-v0.146.0
Key Signal & Impact
Plugin manifests, workspace publishing, Bedrock/Claude Code marketplaces; named sessions, thread fork (incl. temporary); app-server ↔ remote Code Mode over WebSocket
OpenAI Codex CLIrust-v0.146.1
Key Signal & Impact
Backport: safer automatic-review defaults for cyber-capable models; permission changes explained in TUI
Clinev4.1.4
Key Signal & Impact
Breaking plan-mode posture: model can no longer flip plan→act; file-editing shell cmds hard-blocked (not prompt-only); skills shown with workflows; Chutes provider
Clinedesktop-v0.0.8
Key Signal & Impact
Edit any earlier message → fork + workspace checkpoint rewind; transactional/atomic restores; reliable checkpoints after restart/compaction/resume
Clinesdk/v0.0.67
Key Signal & Impact
Reasoning effort/budget normalized from models.dev catalog; Anthropic thinking modes handled explicitly; OpenRouter default → anthropic/claude-sonnet-5
Goosev1.44.0
Key Signal & Impact
Security: arbitrary command execution via goose review / git core.fsmonitor (GHSA-r5pp-p5r8-466r); latest Gemini models; Stop hook gets working_dir
Google Antigravity1.1.10
Key Signal & Impact
Business sign-in for Gemini Enterprise (Cloud project auth, license seats, regional inference, org admin controls)
Gemini CLIv0.54.0
Key Signal & Impact
Enforce HTTPS on GoogleCredentialsAuthProvider (cleartext leak fix); rotate session ID on model fallback; caretaker sanitizes untrusted issue titles
OpenCodev1.18.5
Key Signal & Impact
Claude adaptive thinking + Mistral reasoning/cache stability; correct prompt-cache keys per SDK; server terminal/review transport updates
xAI Grok CLI (Grok Build)ed6d543
Key Signal & Impact
ACP session resume/close; cap live workflow subagents at 16; extract MCP images before truncation; sandbox no longer refuses large deny-glob workspaces
Windsurfv3.6.27
Key Signal & Impact
Devin Desktop Windows loads OS cert store (corporate TLS proxy fix); local edit/write/apply_patch/notebook_edit refuse writes through symlinks
Zed AIv1.15.0-pre
Key Signal & Impact
git.diff_base (HEAD vs default-branch merge base); Project Panel drag-out; JSX/TSX linked editing + Emmet in arrow/return bodies (prerelease)

Theme read: model / agent / cli / mcp / security dominate. Breaking energy clusters on approval & plan-mode boundaries (Codex, Cline) and host trust (Claude self-hosted, Goose fsmonitor, Windsurf symlink writes, Gemini HTTPS).

For developers

  • Codex CLI: Try Agent Plugins and skill import from Cursor/Claude on rust-v0.147.0; if you use auto-approval, prefer --approve-for-me and read the 0.146.1 cyber-model default changes before leaving --full-auto-style flows unsupervised.
  • Cline: On v4.1.4, expect plan mode to refuse mutating shell/file/git/package actions — drive plan→act yourself. On desktop v0.0.8, use mid-thread edit + checkpoint rewind instead of re-prompting from scratch; failed restore should not half-apply.
  • Claude Code v2.1.224: If you need plugins without git/npm, use the archive HTTPS source with SHA-256 pinning; Bedrock users set ANTHROPIC_BEDROCK_REGION_PREFIX when cross-region profiles matter.
  • Goose v1.44.0: Patch immediately if you run CLI goose review or touch git core.fsmonitor — treat GHSA-r5pp-p5r8-466r as actionable, not informational.
  • Gemini CLI v0.54.0: Worth the bump for credential HTTPS enforcement and session-id rotation on model fallback (fewer stateful API errors).
  • Grok Build ed6d543: Note the 16-subagent cap and sandbox behavior change on large deny-glob trees when debugging “sandbox won’t start.”
  • Windsurf v3.6.27: Corporate proxy users on Windows should see sign-in/TLS recover; do not rely on symlink retargeting for approved edits.
  • OpenCode v1.18.5: Practical stability pass for Claude/Mistral thinking and cache keys — upgrade if multi-provider threads were dropping reasoning history.

For teams

  • Enterprise pathing: Claude Code self-hosted runners (Team/Enterprise) and Antigravity 1.1.10 Gemini Enterprise business sign-in are the clearest org-control signals — map them to existing IdP, region, and seat processes before wide rollout.
  • Policy & approvals: Codex’s safer cyber-model auto-review defaults + Cline’s hard plan-mode write blocks are complementary controls. Standardize: who may enable --approve-for-me, and whether plan mode is mandatory for untrusted repos.
  • Supply chain: Prefer Claude’s pinned archive plugins and Codex workspace plugin publishing over ad-hoc marketplace installs; require hash pins and internal catalogs where possible.
  • Incident hygiene: Goose RCE advisory and Windsurf symlink-write refusal are good tabletop prompts — audit agent git hooks, core.fsmonitor, and symlink-heavy monorepos.
  • SDK alignment: Cline SDK v0.0.67 centralizes reasoning budgets via models.dev — teams wrapping Cline should stop hand-rolling per-provider thinking knobs.
  • Open proposals (8) sit mostly at P5 team/developer improve/review for older Codex, Cline, Goose, OpenCode, Zed, and Grok signals — use them as backlog hygiene, not as release blockers for this window.

What to ignore

  • OpenCode pr-38252-videos: verification recordings only — not a product release.
  • Gemini CLI v0.54.0-preview.0 and v0.55.0-nightly.*: noisy prerelease/nightly churn overlapping v0.54.0; favor the stable tag unless you are reproducing a specific fix.
  • Zed AI v1.15.0-pre: real editor signals, but prerelease — don’t set team defaults on it yet.
  • Raw “+N more changes” tails without named behavior: don’t infer features that aren’t in the summaries above.
  • Marketing tone on any HTML changelog; Windsurf’s cert/symlink fixes are concrete — the rest of multi-version page noise is secondary.

Watch next

  • Codex Plugin marketplace + Claude archive/self-hosted runner adoption: will plugin distribution consolidate on pinned HTTPS artifacts?
  • Whether Cline’s “human-only plan→act” and Codex cyber-model approval defaults become the norm across other CLIs.
  • Goose post-GHSA follow-ups (hook/permission surfaces already tagged).
  • Antigravity Enterprise admin-control depth (region, seats, feature gates) in subsequent 1.1.x notes.
  • Grok ACP session resume/close and subagent caps under longer MCP-heavy runs.
  • Resolution of open improve/review proposals on prior Codex/Cline/Zed/Goose cuts so the radar doesn’t double-count stale breaks.

How to use DevAgentRadar

  • Read What the evidence shows first — versioned GitHub/changelog rows only; no vendor mythology.
  • Developers: pick one assistant you already run; apply the concrete flag/mode/security action for that version today.
  • Teams: translate approval, self-host, and advisory items into policy (who can auto-approve, where agents run, what must be pinned).
  • Skip prereleases/nightlies and video-only tags unless you are reproducing a bug.
  • Feed outcomes back via open proposals — evidence in, judgment out, marketing on the floor.

Evidence appendix

  • Evidence window: top curated signals (versioned releases preferred)
  • Assistants tracked: 22
  • Signals in this brief: 17
  • Trusted source mix (GitHub/changelog vs HTML): 55%
  • Open improvement proposals: 8
  • Sources reporting errors: 0

High-signal releases used

  • Claude Code v2.1.224: Added self-hosted environments: claude self-hosted-runner turns your own machines or containers into a place Claude Code web, mobile, and desktop sessions can — source

  • OpenAI Codex CLI rust-v0.147.0: Install portable Agent Plugins and search across local, personal, workspace, and remote plugin catalogs. (#36544, #36409, #36919, #36796) · Organize conversatio — source

  • OpenAI Codex CLI rust-v0.146.0: Name new sessions with /new or /clear, pin important threads, and switch between side conversations without closing them. (#34605, #34840, #35011) · Support — source

  • Zed AI v1.15.0-pre: This week's release includes a new git.diff_base setting for choosing whether editor gutters, file-status colors, and git: diff show uncommitted changes aga — source

  • Google Antigravity 1.1.10: Added Business sign-in for Gemini Enterprise accounts, so you can authenticate with a Google Cloud project under Google Cloud terms, use a license seat allocate — source

  • Cline desktop-v0.0.8: Edit any earlier message in a conversation — the app forks the session at that point, rewinds the workspace to that run's checkpoint, and re-runs from your edit — source

  • Goose v1.44.0: Arbitrary command execution in goose CLI via goose review (via git core.fsmonitor config) - [GHSA-r5pp-p5r8-466r](https://github.com/aaif-goose/goose/security — source

  • Cline v4.1.4: Recognize Chutes as a provider. · Show skills alongside workflows in the slash command menu, and disambiguate commands that share a name instead of letting one — source

  • Cline sdk/sdk/v0.0.67 — SDK v0.0.67: Reasoning controls (effort, budget, on/off) are now driven by the models.dev catalog and normalized once before provider encoding, so requests match what each p — source

  • OpenAI Codex CLI rust-v0.146.1: Apply safer automatic-review defaults for cyber-capable models and explain permission changes in the terminal interface. (#37057) · #37057 [0.146] Backport safe — source

  • Gemini CLI v0.54.0: Changelog for v0.53.0-preview.0 · Changelog for v0.52.0 · chore(release): bump version to 0.54.0-nightly.20260722.gf743ab579 · fix(caretaker): sanitize and wrap — source

  • OpenCode v1.18.5: Improve Claude adaptive thinking handling across more response shapes. · Avoid OpenAI Responses phase handling that could break some conversations. · Preserve g — source

  • Gemini CLI v0.54.0-preview.0: Changelog for v0.53.0-preview.0 · Changelog for v0.52.0 · chore(release): bump version to 0.54.0-nightly.20260722.gf743ab579 · fix(caretaker): sanitize and wrap — source

  • Gemini CLI v0.55.0-nightly.20260729.g3499c84f7: chore/release: bump version to 0.54.0-nightly.20260728.gbef611950 · feat(pr-generator-db): implement Firestore concurrency dual-locking and test ingestion utili — source

Open proposals

Was this guidance useful?

Anonymous · one vote per visitor · re-click to clear. Helps improve digests—not a product ranking.