ceilf6-skills

三视角简历评审：渲染当前简历 PDF，HR 初筛/技术面试官/技术主管三个 persona 并行冷读，产出 problems/questions 各 3 份 + 对账摘要。用户说"三视角评审""跑一轮 persona review""让 HR/面试官/主管看简历"或简历大改后要求重新评审时使用。

Use when the user asks for 三视角评审、persona review、HR/技术面试官/技术主管评审简历，或简历大改后需要重新冷读评审。

Use when Codex needs to create, update, or verify a ByteDance personal daily report in Feishu/Lark wiki for today or a specified date; triggers include 字节日报, 飞书日结, 今日总结, 今天活动记录, bytedance report, lark-cli, bytedcli, Codebase, Bits, Meego, Cloud Ticket, Oncall.

当需要自主循环审视仓库、发现改进点、创建 issue、编写方案、实现变更、提交 PR 并处理 repo-guard 审评反馈时使用。支持 meta-improvement：当 repo-guard 质量偏低时自动优化其 prompts 和 skills。

当需要自主循环审视仓库、发现改进点、创建 issue、编写方案、实现变更、提交 PR 并处理 repo-guard 审评反馈时使用。支持 meta-improvement：当 repo-guard 质量偏低时自动优化其 prompts 和 skills。

当需要对代码变更、Pull Request、自动 CR、合并风险、级联影响、结构质量退化或合入决策进行评审时使用。

Bootstrap an open-source repository Harness from zero to one. Use when Codex needs to design or build agent-ready engineering infrastructure for a new or immature open-source project: contributor docs, governance, issue/PR triage, SDD/TDD workflow, Git hooks, local and CI quality gates, GitNexus impact contracts, release/security automation, repo-guard/Codex/Copilot review loops, and completion verification.

Use when exporting or migrating multiple KM/XueCheng/Citadel documents, especially parent directories, collabpage/page links, daily-report folders, or document lists that need local Markdown files for Notion, docs, archives, or downstream import. Trigger on 批量导出学城, 导出 KM 文档, 学城文章迁移, collabpage 批量, XueChengCopyPlugin, Notion migration from KM.

Use when counting or ranking Friday/SkillHub skills and/or CLI tools created by members of a Daxiang group, especially requests involving group members, group skill statistics, CLI statistics, creator counts, skills per person, Friday skills, Friday CLI, or SkillHub contribution summaries.

纯本地 Semgrep 静态代码分析 Skill。对指定仓库执行全量扫描，查找 Blocker（ERROR）和 Critical（WARNING）级别的安全与质量问题，直接在对话中输出报告。支持 TypeScript、JavaScript、Python、HTML、CSS 等语言。当用户提到静态扫描、代码分析、安全审计、Semgrep、查找 Blocker/Critical 问题、代码质量检查、全量扫描时使用。

读取 code-tape 项目 progress.md，结合项目时间规划、技术模块拆解和工作流程，将 GitHub ID 映射为同学姓名，生成带上下文的进展报告并发送到飞书群聊。

Use when evaluating repo-guard review quality, validating code-reviewer or issue-reviewer prompt changes, comparing reviewer contract changes, or diagnosing degraded PR/Issue comments.

当需要分诊 GitHub issue、评估 issue 质量、分析 bug report 或 feature request 的完整性、清晰度和可执行性，或给报告者提供结构化反馈时使用。

可配置地自动生成并创建个人或团队工作日报/周报。用于写日报、生成日报、今日工作总结、代码提交日报、学城日报、日报创建、本周总结、周报、weekly report 等场景；按 references/config.yaml 中的用户画像、学城目录和群通知配置，日报从学城最近编辑、dev.sankuai.com commit/PR、ONES、TT、日历和用户显式输入中采集当天工作，周报默认基于一周日报聚合工作线和趋势，并通过 citadel 创建、授权或通知。

Design, audit, and refactor permission and safety control planes for AI agents and coding tools. Use when a request involves command or tool approval, permission pipelines, sandbox shortcuts, classifier-assisted adjudication, mode-based safety transitions, headless or remote approval, dangerous allow-rule stripping, or fail-closed operational guardrails. Not for general app auth, browser security, IAM, or unrelated infrastructure hardening.

Design, audit, and refactor the outer CLI entry layer for AI agent tools. Use when a request involves multi-mode CLI design, command dispatch, startup flags, interactive versus headless routing, SSH or remote entry flows, slash-command versus subcommand separation, plugin or skill command registration, command discovery, or cold-start optimization. Not for runtime state-tier design or terminal frame rendering.

Design and audit the runtime state layer for AI agent interfaces. Use when a request involves streaming state, unnecessary re-renders, useSyncExternalStore, external stores, React and non-React state sharing, side-effect centralization, command queues, file watchers, or tiering state by frequency and lifecycle. Not for CLI entry architecture or terminal renderer internals.

Audit and optimize the terminal rendering layer for React-based TUI applications. Use when a request involves terminal flicker, slow frame rate, excessive redraws, high CPU during streaming output, Ink performance, React terminal render pipelines, HostConfig behavior, Yoga layout churn, dirty-flag rendering, screen diffing, or atomic terminal output. Not for browser-only rendering, general app-state design, or CLI entry architecture.

Audit, redesign, and incrementally improve memory systems in agent or coding-assistant projects. Use when Codex needs to inspect how a repo loads context over time, diagnose weak recall or memory bloat, separate durable memory from session memory, add runtime memory surfacing, improve turn-end memory extraction, or design Claude Code-style memory phases across startup preload, in-turn dynamic recall, post-turn persistence, and compaction-time session continuity.