{"id":"39a83f35-e37b-4379-9056-b624ea480e3a","slug":"manuscript-argument-compass","kind":"skill","name":"manuscript-argument-compass","summary":"A cross-platform Agent Skill for keeping one evidence-bounded argument coherent across an academic manuscript.","owner":"chaosbeagle","repoUrl":"https://github.com/chaosbeagle/manuscript-argument-compass","license":"CC-BY-4.0","licenseTier":"permissive","trust":"community","stars":1,"match":null,"fewStars":true,"format":"skill","createdAt":"2026-07-31T08:08:17.000Z","addedAt":"2026-10-21T11:19:54.392Z","logoUrl":"https://avatars.githubusercontent.com/u/32789382?v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["writing","developer-tools"],"tags":["academic-writing","claude-code","codex","research-writing"],"runtime":"binary","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-22T10:20:10.147Z","sample":false,"url":"https://conduits.dev/tools/manuscript-argument-compass","language":"","transports":["stdio"],"install":{},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# Manuscript Argument Compass\n\nManuscript Argument Compass is a reusable Agent Skill for keeping one\nevidence-bounded argument coherent across an academic manuscript.\n\nThe included skill, `steer-question-driven-manuscript`, operates at the\nmeta-writing level. It identifies the paper's latent governing inquiry and\ninternal governing proposition, checks paper identity across sections and\ndisplays, maps claims to evidence, and hands a stable argument compass to\nspecialised drafting or reporting skills.\n\nIt does not replace domain-specific writing, reporting, statistics, citation,\nor venue guidance.\n\n## What it helps with\n\n- diagnosing drift between Introduction, Methods, Results, Discussion, and\n  Conclusion;\n- separating the paper's primary argument from attractive secondary branches;\n- checking whether the design can adjudicate the intended claims;\n- organising Results around the governing inquiry rather than experiment\n  chronology;\n- auditing the reviewer-first path through Title, Abstract, displays,\n  Conclusion, Methods, and Introduction;\n- designing figures and tables as evidence-bearing argument units;\n- keeping alternative explanations distinct from scope or design limitations;\n- calibrating claim strength without making the manuscript defensive.\n\n## Repository layout\n\n```text\nskills/\n└── steer-question-driven-manuscript/\n    ├── SKILL.md\n    ├── agents/\n    │   └── openai.yaml\n    └── references/\n        ├── introduction-argument.md\n        ├── venue-modes.md\n        └── evidence-integrity.md\n```\n\nThe skill content has one canonical source. Codex and Claude Code use the same\n`SKILL.md` and references; only installation and invocation differ.\n\n## Install for Codex\n\nCopy `skills/steer-question-driven-manuscript/` into:\n\n- `$CODEX_HOME/skills/` when `CODEX_HOME` is configured; or\n- `~/.codex/skills/` otherwise.\n\nStart a new task after first installation. Invoke it explicitly with:\n\n```text\nUse $steer-question-driven-manuscript to identify this paper's governing\ninquiry, run the Level 0 identity checksum, and prepare a handoff for the most\nappropriate academic-writing skill.\n```\n\nThe `agents/openai.yaml` file provides Codex/OpenAI interface metadata.\n\n## Install for Claude Code\n\nFor a personal skill available across projects, copy the same directory into:\n\n```text\n~/.claude/skills/steer-question-driven-manuscript/\n```\n\nFor a project-scoped installation, copy it into:\n\n```text\n.claude/skills/steer-question-driven-manuscript/\n```\n\nInvoke it with:\n\n```text\n/steer-question-driven-manuscript Audit this manuscript's governing\nproposition, evidence chain, visual argument, and section-level drift.\n```\n\nClaude Code reads `SKILL.md` and its referenced resources. It does not require\nthe Codex-specific `agents/openai.yaml`.\n\n## Suggested workflow\n\n1. Supply the manuscript or available artifacts, target venue, and article\n   type.\n2. Let the skill reconstruct the latent inquiry rather than forcing a polished\n   research-question sentence.\n3. Review the governing proposition and Level 0 checksum.\n4. Resolve evidence gaps or paper-identity mismatches before sentence editing.\n5. Hand the compass to a specialised writing skill.\n6. Re-audit the finished draft for argument drift and unsupported upgrades.\n\n## Scope and evidence boundary\n\nThe skill can organise supplied evidence and expose contradictions. It must not\ninvent facts, citations, methods, results, uncertainty estimates, or\ngeneralisability. Authors remain responsible for scientific verification and\nthe final manuscript.\n\n## Acknowledgement\n\nThe initial motivation for the Introduction component was influenced by\nShih-Hsin Wang's article, [如何寫出一個有說服力的\nIntroduction](https://shihhsinwang0214.github.io/personal_website/zh/notes/writing-compelling-introduction/).\nThe present repository develops an independently worded, manuscript-wide\nframework. See [NOTICE.md](NOTICE.md) for provenance details.\n\n## License\n\nOriginal material in this repository is licensed under the\n[Creative Commons Attribution 4.0 International License](LICENSE).\n"}