{"id":50733470,"url":"https://github.com/hpsummer/question-to-prompt-pack","last_synced_at":"2026-06-10T11:01:46.114Z","repository":{"id":362391831,"uuid":"1256000185","full_name":"HPSummer/question-to-prompt-pack","owner":"HPSummer","description":"Unified Codex skill for prompt framing, question coaching, user style profiles, and skill routing","archived":false,"fork":false,"pushed_at":"2026-06-04T03:04:41.000Z","size":73,"stargazers_count":2,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-04T04:07:32.576Z","etag":null,"topics":["ai-agents","automation","codex","codex-skills","developer-tools","prompt-engineering","prompt-pack","question-coaching","question-framing","skill-routing","user-intent","user-profile"],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/HPSummer.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-01T11:18:28.000Z","updated_at":"2026-06-04T03:04:35.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/HPSummer/question-to-prompt-pack","commit_stats":null,"previous_names":["hpsummer/question-to-prompt-pack"],"tags_count":6,"template":false,"template_full_name":null,"purl":"pkg:github/HPSummer/question-to-prompt-pack","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HPSummer%2Fquestion-to-prompt-pack","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HPSummer%2Fquestion-to-prompt-pack/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HPSummer%2Fquestion-to-prompt-pack/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HPSummer%2Fquestion-to-prompt-pack/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/HPSummer","download_url":"https://codeload.github.com/HPSummer/question-to-prompt-pack/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HPSummer%2Fquestion-to-prompt-pack/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34149132,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-10T02:00:07.152Z","response_time":89,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-agents","automation","codex","codex-skills","developer-tools","prompt-engineering","prompt-pack","question-coaching","question-framing","skill-routing","user-intent","user-profile"],"created_at":"2026-06-10T11:01:44.682Z","updated_at":"2026-06-10T11:01:46.096Z","avatar_url":"https://github.com/HPSummer.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Question to Prompt Pack\n\n\u003e One unified entry point: understand a rough question, generate the smallest useful prompt pack, then route the task to the right Codex skill when needed.\n\nQuestion to Prompt Pack is a Codex skill for improving user-AI communication. It does not simply make prompts longer. It helps an AI quickly decide whether to answer directly, ask one clarifying question, show a compact collaboration frame, generate a prompt pack, or route the task to the best skill for execution.\n\n中文说明见 [README.zh-CN.md](README.zh-CN.md).\n\n## Quick Demo\n\nUser:\n\n```text\nI want to build a personal research productivity MVP.\n```\n\nQuestion to Prompt Pack:\n\n```text\nI understand this as:\n- Goal: design a small research productivity tool users can build and test quickly\n- Missing/assumed context: assume solo researcher, Codex/Cursor development, notes + tasks + papers\n- Best output: PRD-style MVP plan\n- Mode: tiny planning + route\n\nPrompt pack:\nHelp me design a personal research productivity tool MVP. Focus on the minimum usable workflow for capturing research tasks, linking papers/notes, planning weekly execution, and reviewing progress. Output a PRD-style plan with user stories, core screens, data model, implementation phases, and validation checks.\n\nRoute:\n- Task type: research/planning\n- Best skill: research-execution-copilot\n- Confidence: medium\n- Next action: recommend route, then load selected skill if confirmed\n```\n\n## Why This Exists\n\nMany prompt tools over-expand simple requests. This skill is designed around one rule:\n\n```text\nUse the smallest frame that prevents misunderstanding.\n```\n\nUnified chain:\n\n```text\nrough user question\n-\u003e question-to-prompt-pack aligns intent\n-\u003e concise prompt pack\n-\u003e installed local skills\n-\u003e local discovery cache\n-\u003e first-run GitHub metadata-only discovery\n-\u003e review/install guidance\n-\u003e selected skill executes the task\n-\u003e feedback updates prompt/routing preference\n```\n\nIt helps with:\n\n- turning plain-language questions into structured prompts\n- avoiding overthinking and token waste\n- showing a concise, user-editable interpretation before execution\n- deciding which Codex skill should execute the task\n- teaching one reusable questioning pattern when useful\n- preserving non-sensitive collaboration preferences in a local profile\n- preserving the user's natural style\n- adapting to thread-level preferences through lightweight feedback\n\n## Architecture\n\n```mermaid\nflowchart LR\n  A[\"Rough user question\"] --\u003e B[\"question-to-prompt-pack\u003cbr/\u003eintent frame\"]\n  B --\u003e C[\"Concise prompt pack\"]\n  C --\u003e D[\"Local skill index\"]\n  D --\u003e E{\"Good local match?\"}\n  E -- yes --\u003e F[\"Load selected skill only\"]\n  E -- no --\u003e G[\"Discovery cache\"]\n  G --\u003e H{\"Good cached match?\"}\n  H -- yes --\u003e F\n  H -- no --\u003e I[\"User-approved GitHub\u003cbr/\u003emetadata-only discovery\"]\n  I --\u003e J[\"Review / install guidance\"]\n  J --\u003e K[\"Cache routing record\"]\n  K --\u003e F\n  F --\u003e L[\"Execute task\"]\n  L --\u003e M[\"Lightweight feedback\u003cbr/\u003eprofile / route tuning\"]\n```\n\n## Core Behaviors\n\n| Mode | Use when | Token policy |\n|---|---|---|\n| Tiny Frame | default for ordinary requests | 4 bullets + 1 prompt |\n| Compact Frame | user wants to inspect the AI's understanding | 7 one-line fields |\n| Full Frame | complex task needs assumptions, constraints, and quality criteria | expand only when needed |\n| Training Frame | user wants coaching on how to ask better | diagnosis + exercise + template |\n| Skill Route | specialized workflow should execute the framed task | load one best skill by default |\n| Direct Execution | user says to just do the task | skip framing and execute |\n\n## Question Coaching Loop\n\nWhen the user wants to improve questioning ability, or when a request is missing a high-leverage detail, add a tiny coaching block:\n\n```text\nQuestion upgrade:\n- Missing piece:\n- Why it matters:\n- Reusable pattern:\n```\n\nDefault pattern:\n\n```text\nGoal + context + output format + constraints + execution mode\n```\n\nDo not force coaching into ordinary execution requests.\n\n## Installation\n\nRecommended:\n\n```powershell\n.\\install.ps1\n```\n\nManual install:\n\n```powershell\nCopy-Item -LiteralPath .\\question-to-prompt-pack -Destination \"$env:USERPROFILE\\.codex\\skills\\question-to-prompt-pack\" -Recurse -Force\n```\n\nThen restart or refresh Codex so the skill list is reloaded.\n\n## Usage\n\nUse it as the only front door:\n\n```text\nUse $question-to-prompt-pack:\nUnderstand my rough request, generate a concise prompt pack, choose the best skill if useful, and avoid overthinking.\n```\n\nInitialize a local user style profile:\n\n```powershell\npython .\\question-to-prompt-pack\\scripts\\profile_manager.py --init --validate\n```\n\nBuild a local skill index:\n\n```powershell\npython .\\question-to-prompt-pack\\scripts\\build_local_index.py --out skill-index.json\n```\n\nValidate the unified benchmark:\n\n```powershell\npython .\\question-to-prompt-pack\\scripts\\validate_unified_cases.py --cases .\\benchmarks\\unified-cases.jsonl\n```\n\n## Routing Benchmark Snapshot\n\nThe benchmark currently includes 50 realistic user-style requests across research, coding, writing, PDF/data, image, video, automation, decision-making, and ambiguous inputs.\n\n| Area | Cases | Expected behavior |\n|---|---:|---|\n| Prompt framing | 10 | choose tiny/compact/full/training without over-expansion |\n| Skill routing | 18 | route only when a specialized skill is useful |\n| Direct execution | 8 | skip framing when the request is already clear |\n| Ambiguous/high-risk | 8 | ask one clarification or add verification |\n| Discovery/cache | 6 | use local/cache first, GitHub metadata only after approval |\n\n## Skill Discovery and Routing\n\nDefault routing order:\n\n```text\ninstalled local skills\n-\u003e local discovery cache\n-\u003e first-run GitHub metadata-only discovery\n-\u003e user review/install guidance\n-\u003e later route from local/cache\n```\n\nFirst-run discovery for a new task category:\n\n```powershell\npython .\\question-to-prompt-pack\\scripts\\route_with_discovery.py \"build a React dashboard\" --local-index skill-index.json --discover\n```\n\nDiscovery reads only GitHub `SKILL.md` metadata. It does not auto-install or execute remote code. After the user approves and installs a skill, later requests use the local index or `.question-to-prompt-pack/skill-discovery-cache.json` instead of repeatedly searching GitHub.\n\nConfigure approved discovery sources by copying `sources.example.json` to `.question-to-prompt-pack/sources.json`:\n\n```json\n{\n  \"refresh_policy\": \"weekly\",\n  \"sources\": [\n    {\n      \"name\": \"openai-skills\",\n      \"url\": \"https://github.com/openai/skills\",\n      \"enabled\": true,\n      \"trust_level\": \"review\"\n    }\n  ]\n}\n```\n\n## Repository Layout\n\n```text\nquestion-to-prompt-pack/\n  SKILL.md\n  agents/openai.yaml\n  references/\n  assets/\n  scripts/\nbenchmarks/\n  unified-cases.jsonl\nexamples/\n  before-after.md\n```\n\n## License\n\nMIT. See [LICENSE](LICENSE).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhpsummer%2Fquestion-to-prompt-pack","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhpsummer%2Fquestion-to-prompt-pack","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhpsummer%2Fquestion-to-prompt-pack/lists"}