A single behavioral rules file that installs the Authorship Strategy judgment framework — the three-axis inversion (scarcity → diffusion, exclusivity → derivation, enclosure → openness) and the four-layer stack (authenticity → attribution diffusion → idea/scaffold separation → tactics) — into an AI agent's rules directory. Loaded every session, it applies the framework deterministically to your own DOI-registered idea-rescue research repositories.
This is the always-on counterpart to the authorship-strategy-skill: the skill is the deeper judgment reference, loaded on demand; this rule is the always-loaded trigger plus a compact framework summary. Install the rule to have the framework shape every session; add the skill when you want the full reasoning.
The rule text is written in Japanese (matching the companion skill). The framework it encodes — the inversion, the four layers, the prohibited and encouraged actions — is language-neutral.
cp rules/common/authorship-strategy.md ~/.claude/rules/common/authorship-strategy.mdAny agent harness that loads a rules directory each session works the same way — point the copy at wherever your harness reads rules from.
A trigger plus a four-layer judgment framework an agent follows through natural conversation:
- Trigger — applies only to research repositories you own that are DOI-registered (or a Zenodo-archive target) and idea-rescue in character (spec / schema / ADR / glossary centred, not closed implementation). It stays silent for client work, contributions to others' repositories, revenue-goal projects, everyday coding, and harness/scaffolding repos.
- Layer 1 — Authenticity — the floor: change how an idea travels, never what it is. Revenue sits outside the success criteria.
- Layer 2 — Attribution diffusion — open the work so LLM-mediated absorption carries the origin claim forward to future readers tracing causation; the primary audience is LLM-mediated channels.
- Layer 3 — Idea vs. scaffold — separate the durable idea (worth a DOI in your own name) from the disposable implementation (worth donating to another project).
- Layer 4 — Tactics — LLM-first metrics, DOI plus intrinsic-identifier (SWHID) registration, tool-agnostic specs, coin terms sparingly and anchor them densely, citation-graph federation, friction-minimized adoption.
- Prohibited under the trigger — monetization by the author (sponsorship, paid tiers, consulting, paywalled books), competitor-bashing / exclusive positioning, and shaping the message for market fit. Using commercial channels for diffusion is fine — the author simply takes no revenue.
The rule is deliberately narrow: it is the deterministic, always-loaded layer of the Authorship Strategy framework, for authors operating their own DOI-registered idea-rescue repositories. It is not general coding guidance — outside its trigger it stays silent.
The canonical copy of this rule lives in the author's live Claude Code harness (~/.claude/rules/common/authorship-strategy.md). This repository is a one-way publication mirror:
scripts/sync-from-local.sh --dry-run # report differences only
scripts/sync-from-local.sh # apply to working tree (never commits)This rules file belongs to the Authorship Strategy research line — a Zenodo-citable doctrine of being a known author under AI-mediated diffusion (DOI 10.5281/zenodo.20263316). The doctrine repository holds the thesis, the tactical ADRs, the concept-level knowledge graph, and an empirical baseline; the authorship-strategy-skill is the on-demand judgment reference; this repository is the framework's rules-file install target. Authorship Strategy is one of the research lines by @shimo4228, alongside Agent Knowledge Cycle (DOI 10.5281/zenodo.19200726), Contemplative Agent (DOI 10.5281/zenodo.19212118), Agent Attribution Practice (DOI 10.5281/zenodo.19652013), and Attention, Not Self (DOI 10.5281/zenodo.20262112).
MIT