Diagnose and harden Codex workflows against common user-reported failure modes: context compaction drift, forgotten AGENTS.md instructions, missing long-term memory, repeated approvals, slow or stuck turns, tool_suggest overhead, and unclear session state.…
Organize messy, long, or resumed conversations into concise reusable context. Use when the user asks to tidy,整理,梳理,压缩,总结,归档,交接, or迁移 a conversation; when extracting decisions, actions, artifacts, sources, open questions, or next resume prompts; or when…
Mine conversation archives, topic memory, idea inboxes, and project notes for reusable agent-skill candidates, then classify which should be packaged, reviewed, or kept private. Use when Codex is asked to search chats/conversations for possible skills, turn…
Audit GitHub repositories that publish agent skills and summarize real adoption signals: stars, forks, watchers, issues, recent views, clones, referrers, popular paths, directory/marketplace indexing, and external references. Use when Codex is asked whether a…
Turn a research paper into a structured, styled PowerPoint deck: read a lawful paper/PDF, reconstruct the paper logic, create a top-level presentation outline, generate slide-level titles, map each slide to paper text/figures/tables/equations, create…
Write a source-grounded research paper through a legal literature and drafting pipeline: study writing-method sources, search lawful paper indexes and open-access databases, build section/paragraph structures, generate original figures/tables/equations from…
Run a design thinking workflow: empathize, define, ideate, prototype, and test. Use when Codex is asked to solve user-centered product, service, interface, education, or experience problems through research, framing, ideation, prototyping, and user feedback.
Run a DMAIC process improvement project: define, measure, analyze, improve, and control. Use when Codex is asked to fix an existing process with measurable defects, delays, variation, waste, quality problems, or customer-impacting failures.