v1.0.8.0 — SpecKit Dual-Threshold & Five Checks Framework

This release updates the OpenCode environment with work on dual-threshold validation (gate 2), five checks framework, script reorganization. The goal is to make the platform clearer, more reliable, and easier to use day to day.

Dual-Threshold Validation (Gate 2)

• Need: This part of the release makes agent behavior more consistent across the supported runtimes.
• What changed: READINESS Formula: (confidence >= 0.70) AND (uncertainty <= 0.35) - Confidence alone is insufficient; now requires both dimensions. "Confident Ignorance" Detection: High confidence + high uncertainty triggers investigation protocol (max 3 iterations before escalate). 4-factor Uncertainty Model: Epistemic gaps (0.30), Model boundaries (0.25), Temporal variability (0.20), Situational completeness (0.25).
• Why it matters: That matters because the same instructions now behave more predictably in different runtimes.

Five Checks Framework

• Need: This part of the release improves five checks framework.
• What changed: Architectural Validation: Required for Level 3/3+ specs, recommended for Level 2. 5 Checks: Necessary? (solving actual need), Beyond Local Maxima? (alternatives explored), Sufficient? (simplest approach), Fits Goal? (on critical path), Open Horizons? (long-term aligned). New Reference: references/validation/five-checks.md.
• Why it matters: That matters because it makes the environment easier to use and maintain.

Script Reorganization

• Need: This part of the release improves how the environment saves, finds, and uses context.
• What changed: Decision-tree-generator.js: Moved extractors/ → lib/ (generator, not extractor). Opencode-capture.js: Moved lib/ → extractors/ (captures session data). Import Paths: Updated in 5 files (decision-extractor, diagram-extractor, data-loader, index.js, test file).
• Why it matters: That matters because saved context is easier to find, easier to trust, and less likely to get in the way.

Documentation Overhaul

• Need: This part of the release improves how the environment saves, finds, and uses context.
• What changed: AGENTS.md: Agent Routing now Section 6, Tool System → 7, Skills System → 8. README.md: Streamlined with -625 net lines reduction. New References: epistemic-vectors.md, decision-format.md, five-checks.md.
• Why it matters: That matters because saved context is easier to find, easier to trust, and less likely to get in the way.

Fixed

• Need: This part of the release improves how the environment saves, finds, and uses context.
• What changed: Gate Numbering: Consistent sequential numbering throughout documentation. Gate-enforcement.md Line 199: "Gate 3" → "Gate 1" for Memory Context Loading. SKILL.md Line 855: "4 files + README" → "5 subdirs + README" for templates/verbose/.
• Why it matters: That matters because saved context is easier to find, easier to trust, and less likely to get in the way.

---

Files Changed

File or Area	Note
4 core files (AGENTS.md, README.md, CHANGELOG.md, skill_advisor.py)	Updated in this release.
30 system-spec-kit files: SKILL.md, mcp_server/package.json, constitutional/gate-enforcement.md, scripts/extractors/ (6 files), scripts/lib/ (2 files), scripts/loaders/ (1 file), references/ (4 new + 4 modified), templates/ (3 files), mcp_server/handlers/ (2 files), mcp_server/README.md	Updated in this release.

---

Upgrade

• No action required. Pull latest to get dual-threshold validation in skill routing, Five Checks Framework for architectural decisions, and reorganized scripts with correct import paths.

Full changelog: .opencode/changelog/