// Track

Agent-Delegated Coding

Specs that eliminate clarification round-trips

How to write sub-agent specs that return working code on the first try. File paths, line numbers, scaling factors, acceptance criteria, backtest methodology — the anatomy of a gold-standard spec, drawn from the Hermes PR #28 delegation that took a scoring rebalance from concept to merged PR with zero back-and-forth.

5 lessons~37 min total

Lessons are shown in recommended order. Complete them in sequence for the best experience — or jump to any lesson.

Lesson 273·9 min read

Anatomy of a Gold-Standard Spec

The Hermes PR #28 spec breakdown. File paths, line numbers, exact scaling factors, acceptance criteria, backtest method. Zero clarification round-trips. Why specificity beats narrative when dispatching to sub-agents — and what a template looks like.

Lesson 274·7 min read

File Paths, Line Numbers, and Scaling Factors

Precision in the spec eliminates clarification round-trips. The difference between 'rebalance the scoring' (bad) and 'change grok_feed.py:274 from deviation * 30.0 to deviation * 35.0' (good) is the difference between three attempts and one.

Lesson 275·7 min read

Acceptance Criteria as Backstops

Defining 'done' before the agent starts. The regression check rule — previously-passing signals must still pass. Target ranges for backtest hit rates. Why acceptance criteria are the difference between 'looks plausible' and 'is actually correct'.

Lesson 276·7 min read

The Review Loop — CodeRabbit Catches What Agents Miss

The calibrator continuity bug is a real example. The code-writing agent did not catch the 18.4 to 17.6 discontinuity at the 0.15 boundary. CodeRabbit did. When to trust the agent, when to verify manually, and why automated reviewers are non-optional for math-heavy changes.

Lesson 277·7 min read

When Agents Doom-Loop

The session retro pattern: three doom-loops on signal_engine.py in one session. Each time the agent produced a 'fix' that broke a different thing. How to detect doom-loops early, intervene, and recover without burning another two hours.