feat(prompts): inject code-minimalism discipline into codegen

[ceilf6]

Linked Issue Or Context

Closes #367.

Summary

• Adds a CODE_MINIMALISM_DISCIPLINE constant to packages/core/src/llm/prompts.ts, parallel to the existing PROGRESSIVE_EXPLORATION_PROTOCOL, EXTERNAL_KNOWLEDGE_PROTOCOL, and SECURITY_DISCIPLINE.
• Injects it into both executor codegen system prompts (generateCodeForFile and generateModifiedCode) so generated/modified code avoids the common AI over-engineering failure modes: speculative/premature abstractions, defensive code for impossible scenarios (validate only at system boundaries), gratuitous comments, and scope creep beyond the task.
• Distinct from existing lines: 只修改必要部分 constrains the scope of a change; 保持原有代码风格 constrains style matching; neither constrains internal over-engineering. generateCodeForFile (new files) previously had no minimalism guidance at all.
• Adds focused prompt-capture tests with one distinct anchor phrase per sub-discipline (per the feat(prompts): inject code-minimalism (anti-gold-plating) discipline into codegen prompts (learned from Claude Code agentic prompt) #367 repo-guard suggestion) to keep assertions robust to prompt-copy tweaks.

Learned from the Claude Code agentic prompt's "Doing tasks" minimalism bullets (Claude-Code-source/src/constants/prompts.ts:201-203, 207-208).

Impact Scope

Prompt-only change in packages/core/src/llm/prompts.ts (new constant) and packages/core/src/llm/code-generation.ts (two system-prompt injections), plus focused tests. No schema, tool routing, executor contract, or planner changes.

GitNexus Impact Summary

• Risk level: LOW
• Critical skeleton changes: none
• GitNexus impact: impact(generateCodeForFile, upstream) and impact(generateModifiedCode, upstream) both returned LOW with 0 affected processes and 0 affected modules; system-prompt string append only, signatures/return contracts unchanged.
• Verification: gitnexus detect_changes (scope=unstaged) reported affected_count=0, risk_level=low, zero affected processes; changed symbols limited to the two codegen functions and test helpers.

Verification

• pnpm --filter @frontagent/core test -- code-generation.test.ts — 13/13 pass (6 system-prompt tests: 2 external-knowledge + 2 security + 2 minimalism).
• pnpm exec biome check on the 3 changed files — clean.
• pnpm quality:precommit (lint + typecheck + test + test:workflows) — green (25 turbo tasks successful, workflows 32/32).
• gitnexus impact (both functions) + gitnexus detect_changes — LOW, 0 affected processes.

Checklist

• I have linked an issue or explained why this PR stands alone.
• I have kept the diff focused on the stated change.
• I have run pnpm quality:precommit, or explained why it could not run.
• I have run pnpm quality:local for critical skeleton changes, or explained why it could not run. (Prompt-only, non-critical; precommit + GitNexus impact/detect_changes cover it.)
• I have updated docs or tests when behavior, public APIs, or Harness contracts changed.
• For critical skeleton changes, I have filled the GitNexus impact summary with concrete results.

[github-actions]

🛡️ ceilf6/repo-guard

代码评审报告: feat(prompts): inject code-minimalism discipline into codegen

风险等级: 中
处理建议: 批准
决策摘要: ** 可合并；变更满足 #367 的 prompt-only 验收标准，未改变签名/schema/调用契约，新增测试覆盖了两条 codegen system prompt 的注入路径。

级联分析

• 变更符号: 原始模型未提供结构化级联字段。
• 受影响流程: 原始模型未提供结构化级联字段。
• 变更集外调用方: unknown
• 置信度: degraded

问题发现

模型未返回可结构化的问题发现；已提取可用的决策字段，原始非契约内容未附在评论中。

行级发现

• 无明确变更行归属。

Karpathy 评审

• 假设: 模型输出需要归一化为固定 Markdown 契约。
• 简洁性: 已提取 summary、finding、evidence 与 fix；原始 prose 不再附在评论中，避免占用下游解析与代理上下文。
• 变更范围: 原始模型未提供结构化范围字段。
• 验证: 需要查看 CI、测试或人工 CR 证据补强合并信心。

缺失覆盖

• 输出未命中 Repo Guard Markdown 契约；建议补充真实模型质量评估覆盖。