All issues

Search results

Add minimum-sufficient execution policy and trajectory evals

Status: Open.

#135 In JRichlen/agent-plugins;

· JRichlen opened on Sep 13, 2026

Freeze a routing holdout that may never be regraded

Status: Open.

#126 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Behavioral coverage: 5 packs have no calibration control, 13 plugins have no pack at all

Status: Open.

#125 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Cross-harness is the marketplace's central claim and is tested for 2 of 25 plugins, on 1 harness in CI

Status: Open.

#124 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Build a trace corpus and derive eval scenarios from observed failures, not from design intent

Status: Open.

#123 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Validate the graders: report TPR/TNR, not raw agreement, and take the first measurement

Status: Open.

#122 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Harden safety-path grading: pass^k floor for irreversible actions, and bound the pier verifier's collateral damage

Status: Open.

#121 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Eval strategy assessment: coverage, gaps and recommendations across every tier

Status: Open.

#119 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

Adaptive behavior v1: learn from feedback and prove one useful improvement

Status: Open.

#117 In JRichlen/agent-plugins;

· JRichlen opened on Sep 8, 2026

[Aperture 6/9] Define a portable agent request metadata and capability contract

Status: Open.

#101 In JRichlen/agent-plugins;

· JRichlen opened on Sep 5, 2026