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: Qwen Code Proposal: Import Vitest 3.2 JSON reports into Qwen Code Test Failure Explorer Summary: Qwen Code v0.25.2 imports Vitest 3.2 JSON reports via MCP tools and a Test Failure Explorer that searches, pages, and copies analysis requests for up to 20 failures. Pair that with --disable-mcp-server, H3 background Shell, and opt-in TypeScript headless Mods. Primary source: https://github.com/QwenLM/qwen-code/releases/tag/v0.25.2

Why it matters

v0.25.2 has no known breaking changes. The highest-leverage workflow change is the test-failure loop: a private extension imports Vitest 3.2 JSON reports and exposes failures, errors, counts, and diagnostics through MCP tools with search and pagination, while the Test Failure Explorer browses, searches, pages, and copies analysis requests for up to 20 selected failures. Paste those copied requests into the agent instead of dumping raw logs. Use repeatable --disable-mcp-server to drop unused MCP servers per invocation; the CLI now strips a leading UTF-8 BOM from project .mcp.json and --mcp-config. Context-overflow retries use the server-reported context limit, so reactive compaction recovers more reliably. Enable H3 Monitor and background Shell so runs outlive their turn and still write artifacts. Treat Monitor admission and initial-call startup failures as tool errors with guidance, not success. Opt-in headless Mods add TypeScript authoring, relative imports, typed non-secret configuration, pre-startup dependency validation, ordered middleware, command interception, and bounded failure recovery without resetting module state. In trusted headless main sessions they can intercept tool calls, rewrite arguments, refuse calls, return synthetic results, and check permissions. Keep foreground Shell on authorized public REST/WebShell sessions default-off unless you need mandatory-approval Shell.

Suggested actions

1. Upgrade Qwen Code to v0.25.2 from https://github.com/QwenLM/qwen-code/releases/tag/v0.25.2 (no known breaking changes).
2. After Vitest 3.2, point Qwen Code at the JSON report and instruct: Import this Vitest 3.2 JSON report via the private MCP extension; expose failures, errors, counts, and diagnostics; search and paginate until every failing test is listed.
3. Open the Test Failure Explorer, browse/search/page the imported failures, select up to 20 tests, copy the generated analysis requests, and paste them into the current agent session as a single fix batch.
4. Launch Qwen Code with repeatable --disable-mcp-server <server-name> flags to exclude unused MCP servers for that invocation; keep project .mcp.json and --mcp-config inputs free of a UTF-8 BOM (the CLI now strips one if present).
5. Enable H3 Monitor and background Shell so commands outlive the turn and write artifacts; on trusted headless main sessions, opt in TypeScript Mods (relative imports, typed non-secret config, pre-startup dependency validation) to intercept, rewrite, refuse, or synthesize tool results with ordered middleware and bounded failure recovery—without resetting module state.

Config surfaces this release may change

• MCP servers — check your repo before applying
• settings files — 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/qwen-code-v0-25-2-import-vitest-3-2-json-reports-into-qwen-code-test-fai and mark Applied, Skipped, or Failed. Proposal id: 45a45d0e-a88d-466c-a8ff-1bb5694fc3a5

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.