DevAgentRadar → Repo Harness Rule Patch

Goal: update our repository's permanent harness instructions based on this release.

Update the file that tool actually reads. Do not dump everything into one file.

• Claude Code → CLAUDE.md plus .claude/ (skills, hooks, settings, agents, commands). It does not read AGENTS.md natively; start CLAUDE.md with @AGENTS.md.
• Codex, Copilot, Cursor, Factory Droid, Grok Build, Roo Code, Goose, OpenCode, Amp, Zed, Aider → AGENTS.md.
• Gemini CLI / Antigravity → GEMINI.md. AGENTS.md only if context.fileName is set.
• Cursor glob-scoped rules → .cursor/rules/*.mdc (plain .md is ignored), not a second constitution.
• Codex MCP → .codex/config.toml, not .mcp.json.
• Copilot extra instructions → .github/copilot-instructions.md; custom agents → .github/agents/.
• Windsurf → .windsurf/rules (do not assume AGENTS.md).
• Cline → .clinerules.
• Procedures → a skill (SKILL.md). Enforcement the model must not skip → a hook. Isolated roles → subagents.
• Do not fork the same rule into three tool files.

Context

Assistant: Zed AI Proposal: Enable effort-based reasoning for OpenRouter models in Zed Agent Summary: Zed 1.19.0-pre adds OpenRouter effort-based reasoning, lower memory for long Agent sessions, Terminal Thread renaming, call hierarchy (cmd-k cmd-h / ctrl-k ctrl-h), and Git Panel multi-select. Keep threads alive, feed caller graphs to the agent, and batch Git actions. Primary source: https://github.com/zed-industries/zed/releases/tag/v1.19.0-pre

Why it matters

Zed v1.19.0-pre is a prerelease whose Agent and navigation changes should reshape how you loop in the editor. OpenRouter models now support effort-based reasoning, so spend more compute on hard refactors without leaving Zed. Long Agent sessions use less memory, OpenAI-compatible streaming is improved, the UI no longer hangs when providers stream large tool inputs, macOS Agent terminal tool calls no longer leak file descriptors, and Anthropic HTTP 400 "prompt is too long" errors are recognized as context-window overflows—so keep a thread alive instead of bouncing. Rename Terminal Threads from the Threads Sidebar so parallel agent terminals stay identifiable. Before a refactor, run call hierarchy (show incoming/outgoing calls; cmd-k cmd-h on macOS, ctrl-k ctrl-h on Linux/Windows) and use Git Panel multi-select to gather callers and file sets yourself rather than waiting for the agent to grep. Untitled buffers get automatic language detection, so scratch prompts and snippets highlight without a manual language pick. Do not assume anything from the 18 omitted changelog entries.

Suggested actions

1. In Zed Agent, select an OpenRouter model and enable effort-based reasoning (added in v1.19.0-pre).
2. Keep one Agent thread for the full task instead of restarting: long sessions use less memory, OpenAI-compatible streaming is improved, large tool-input streams no longer hang the UI, and Anthropic HTTP 400 "prompt is too long" is treated as a context-window overflow.
3. Open the Threads Sidebar and rename each Terminal Thread to the task it belongs to.
4. From the current function, run command palette call hierarchy: show incoming calls then call hierarchy: show outgoing calls (macOS: cmd-k cmd-h; Linux/Windows: ctrl-k ctrl-h) and paste the callers/callees into the Agent prompt before asking for a refactor.
5. In the Git Panel, multi-select related files and act on them as a group instead of sending the Agent a one-file-at-a-time git loop.

Config surfaces this release may change

• MCP servers — check your repo before applying

After you finish

Do not report this as applied to DevAgentRadar. You cannot write the visitor's loop.

Tell the human: open https://devagentradar.com/proposals/zed-ai-v1-19-0-pre-enable-effort-based-reasoning-for-openrouter-models-i and mark Applied, Skipped, or Failed. Proposal id: b97c1537-b9e6-4564-bb21-106712536ec6

Instructions:

1. Read the existing context files and settings for the tools in this repo.
2. Draft an explicit Git diff. Update ONLY agent harness rules or tool configuration.
3. Call out deprecated flags, obsolete habits, or changed permission boundaries.
4. Do not touch application logic.

Start with the proposed diff and say which layer and which tool's file it belongs in.