project: hivespec plugin — tracking issue #832
Description
Summary
Tracking issue for HiveSpec — phase-based delivery lifecycle for AI agent swarms. Eval results at HiveSpec-Evals.
Status
Plugin implemented and published as standalone repo. Evals pass on pi-cli (10/17, 0.824 mean). Claude-cli blocked by #830.
Sub-Issues
Done
- feat: add hivespec plugin (phase-based delivery lifecycle for agent swarms) #823 — Initial plugin implementation
- feat: improve eval authoring guidance based on autopilot-dev learnings #829 — Eval authoring learnings (decomposed into docs: document that workspace setup must copy skills to .claude/, .agents/, and .pi/ #834-feat: support directory paths in eval run command #837)
Ready for handoff
- bug: target segment not removed from artifact paths despite PR #801 #827 — Remove target segment from artifact paths (one-line fix, exact file+line provided)
- bug: claude-cli provider crashes after ~4 tests in combined eval run (SessionStart hook error) #830 — Claude-cli provider crashes after ~4 tests (repro steps + investigation pointers)
- docs: document that workspace setup must copy skills to .claude/, .agents/, and .pi/ #834 — Document workspace setup must copy to .claude/, .agents/, .pi/
- docs: document that workspace evals should not depend on GitHub remote #835 — Document workspace evals should not depend on GitHub remote
- docs: document that agents verify git diff against prompt claims #836 — Document agents verify git diff against prompt claims
- feat: support directory paths in eval run command #837 — Support directory paths in eval run command
- bug: dataset name includes .eval suffix from filename (hs-design.eval instead of hs-design) #838 — Dataset name includes .eval suffix from filename
Needs design first
- feat: auto-push eval results to configurable git repo (needs design) #826 — Auto-push eval results to configurable git repo
Future
- Register hivespec.dev domain
- Linear/Jira adapter for hs-claim (currently GitHub-only in examples)
- OpenSpec alignment (GIVEN/WHEN/THEN specs, verify-report.md, archive phase)
- Remove SKILL.md file inputs from evals once workspace skill discovery is reliable (docs: document that workspace setup must copy skills to .claude/, .agents/, and .pi/ #834)
Related
- EntityProcess/HiveSpec — Plugin source
- EntityProcess/HiveSpec-Evals — Eval result artifacts