Running subagents across harnesses: a five-part independent code review workflow
PawelHuryn · x · 2026-11-21
Pawel Huryn shares his full setup for coordinating coding subagents across harnesses: Claude Code (Opus 5.5 at xhigh) coordinates and handles small fixes, Codex on GPT-6.2 Sol builds bigger changes, and every review runs in a fresh session—never reusing the thread that built the code.
His five-part reviewer brief, in order: (1) what the product is in two sentences; (2) the change straight from git, never a written summary; (3) who uses what, how often, and what's serious even when rare; (4) "what would block or break this release?" with each finding labeled INTRODUCED or PRE-EXISTING plus exact repro steps; (5) a ranked list of claim, file:line, and fix.
Key discipline: no earlier findings, suspected files, or "do you approve?"—ask a reviewer if it approves and it approves. Claude Code verifies every finding against the code before fixes. Both skills (codex-agents, independent-review) are open-sourced free.
More from coding & agent
- Vespa co-founders launch Hornet: solving retrieval for the agentic era — jobergum · 2026-11-21
- New Company Agent Bench: Gemini Neon leads at 87.4%, simple API harnesses beat products — philhchen · 2026-11-21
- Head-to-head coding test: Tencent Hy4 Flash finishes in 7:38 vs DeepSeek's 90 minutes — HeyToha · 2026-11-21
- Antigravity's new /teamwork feature lets you assign tasks to multiple agents at once — venturetwins · 2026-11-21
- Zed CEO mocks AI PR spam: opening 100 PRs to fix one bug isn't the flex you think — zeeg · 2026-11-21
- Cohere researcher talks rethinking coding-agent evaluation at Agent Eval Symposium — nouhadziri · 2026-11-21