v10.28 strategic correction posture hardening follow-up

Summary

• make correction reasoning start from shared logic before local symptoms
• make fix scope an evidence-earned conclusion rather than a convenience-first supplier/model/path assumption
• update the related playground case and governing master surfaces for the P120 follow-up wave

Verification

• installed active rules into `$HOME/.claude/rules`
• verified 18/18 source/runtime parity and source/destination body sufficiency
• ran `git diff --check`

v10.27 active-exchange language alignment hardening follow-up

Summary

• make goal-shaped and recommendation-shaped scaffold follow the user's active exchange language by default
• keep explicit language requests as a stronger override while preserving exact literals token-by-token
• update the related playground case and governing master surfaces for the P119 follow-up wave

Verification

• installed active rules into `$HOME/.claude/rules`
• verified 18/18 source/runtime parity and source/destination body sufficiency
• ran `git diff --check`

v10.26 successor-surfacing bridge hardening follow-up

Summary

• harden successor-surfacing bridge behavior so visible next work is not left as a generic future note
• derive smaller bounded successor slices from governed execution surfaces when upstream labels stay too broad
• update the related playground case and governing master surfaces for the P118 follow-up wave

Verification

• installed active rules into `$HOME/.claude/rules`
• verified 18/18 source/runtime parity and source/destination body sufficiency
• ran `git diff --check`

v10.25: proactive goal surfacing and response style

• make candidate-goal surfacing more proactive at real decision boundaries when several next slices remain live
• refine default non-trivial answer style to be easy-first, compact-but-complete, scanable, identifier-meaning-first, evidence-layer-clear, and decision-ready
• add playground case 17 for proactive goal surfacing and decision-ready explanation while keeping playground non-runtime

v10.24: end-to-end language-aligned goal surface

• refine RULES so candidate goals, advisory /goal, recommendation labels, and recap/closing lines follow the dominant session language end-to-end
• preserve exact literals such as /goal, file paths, version tags, and code identifiers when they should remain exact
• add playground case 16 for end-to-end language-aligned goal-surface behavior while keeping playground non-runtime

v10.23: language-aware candidate-goal promotion playground case

Summary

• add one new governed non-runtime playground case for language-aware candidate-goal promotion
• index the new case in the playground README
• keep the playground outside the active 18-file runtime install payload

Verification

• reinstalled the active runtime rules into ~/.claude/rules
• verified 18/18 source/runtime parity plus source/destination body sufficiency
• ran git diff --check clean

v10.22: language-aware candidate goal and /goal doctrine

Summary

• make candidate goals and promoted /goal suggestions follow the dominant session language by default
• shape multi-path successor recommendations as candidate goals before governed command promotion
• promote only the best-supported governed repo candidate into advisory /goal form under the existing governed-work-only boundary

Verification

• installed the active runtime rules into ~/.claude/rules
• verified 18/18 source/runtime parity plus source/destination body sufficiency
• ran git diff --check clean

v10.21: governed-work-only /goal context sourcing doctrine

Summary

• tighten /goal so trivial non-governed next steps stay concise by default
• require design-first governed-surface sourcing for bounded repo-governed successor work
• include changelog, patch, and README only when they materially shape completion, review, or current-state impact

Verification

• installed the active runtime rules into ~/.claude/rules
• verified 18/18 source/runtime parity plus source/destination body sufficiency
• ran git diff --check clean

v10.20

Summary

• upgrade playground/ with transcript-grounded observed cases from real Claude Code session JSONL transcripts on this machine
• add richer multi-turn scenario traces, stronger flow explanations, and two grounded new scenario families
• preserve the 18-rule runtime install boundary while keeping playground/ outside the runtime payload

v10.19 - governed RULES playground scenarios

Summary

• add a governed non-runtime playground/ family for showing how current RULES change AI behavior in practice
• add 10 scenario-family case files with prompt/response examples and lightweight flow diagrams
• add full 18-rule coverage mapping, a virtual-case matrix, an observed-case log, and a reusable case template
• keep README integration pointer-level only and preserve the 18-rule runtime install boundary

Verification

• 10 scenario-family files created under playground/cases/
• every scenario separates rule-enforced fact, observed case, and virtual variant
• every scenario includes example dialogue and flow diagram sections
• coverage matrix maps all 18 active runtime rules to at least one scenario family
• virtual matrix includes request type, evidence state, scope clarity, risk level, and expected rule response axes
• runtime install still verifies 18/18 parity and keeps playground/ out of .claude/rules/
• git diff --check clean in source scope