All notable changes to Agent-PROVE.
- Gemini CLI support —
gemini-extension.json+GEMINI.mdwith tool mapping and subagent fallback note - 4 new commands —
/brainstorm,/consider,/review,/frameworkscompleting the 7-command user journey - 3 new skills — brainstorm, consider, review (matching command workflows)
- ADR-005 — Command journey design and phase-aware defaults rationale
- LICENSE — MIT license file
- CONTRIBUTING.md — Guide for adding custom frameworks with template
- Session handoff —
.planning/HANDOFF.mdfor context continuity
- README.md — Full polish with badges, structured sections, platform table, all 7 commands
- PRODUCT.md — Updated with 7-command journey and "Won't Do" positioning
- 5 new framework agents — six-thinking-hats, gap-finder, blind-spot, devils-advocate, drift-detector
- Phase-aware defaults —
/validateauto-selects Exploration + Deconstruction + Strategy;/thinkauto-selects Verification + Agentic Audit .claude-plugin/plugin.json— Marketplace-standard manifest.claude-plugin/hooks.json— SessionStart hook announcing plugin- Multi-platform configs —
.cursor-plugin/,.codex/,.opencode/ examples/— 3 real-world output examples (validate, think, audit)scripts/validate-structure.sh— Plugin structure validation- Research 002 — New agent value validation and style pattern analysis
- PRODUCT.md — Marketing-grade product description with competitive positioning
- CLAUDE.md — Added framework taxonomy table with phase-aware defaults
- README.md — Updated to 14 frameworks with Classic and Agentic Era sections
- T1 Validator — Updated dispatch list to include all 14 frameworks
- Skills — validate-approach and think-cycle updated with phase-aware framework defaults
- 9 framework agents — first-principles, inversion, five-whys, occams-razor, via-negativa, one-thing, deep-dive, technical-research, tri-strike
- 2 orchestrator agents — T1 validator (pre-execution) and cycle auditor (per-cycle)
- 3 skills — validate-approach, think-cycle, evidence-audit
- 3 commands —
/validate,/think,/audit - Evidence protocol — Zero uncited inference enforcement with violation taxonomy
CLAUDE.md— Plugin architecture and conventionsREADME.md— User documentation- ADRs 001-004 — Prior art, architecture, evidence protocol, agent protocol
- Research 001 — Competitive landscape analysis (13 repos, 12 installed plugins)
.planning/— Full planning artifacts (brief, roadmap, 8 phase plans)