一键在 Manus 中运行任何 Skill

migrate-research

星标30

分支0

更新时间2026年4月8日 20:42

Autonomous research loop that improves citum-migrate converter fidelity by iterating on Rust code in crates/citum_migrate/src/. Measures fidelity across a corpus of priority styles, clusters failure patterns, attempts targeted Rust fixes, validates with cargo nextest, keeps improvements, reverts failures. Does NOT modify style YAML.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

citum

citum/citum-core

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

Migrate Research

Use This Skill When

Systematic converter improvement is the goal.
Multiple styles share the same migration failure pattern.
You want to close the gap between citum-migrate output and reference fidelity.
A style-fidelity follow-up needs bounded rich-input evidence to determine whether the next pass belongs in migration, style, processor, or adjudication.

Do Not Use When

Fixing a single style's YAML.
The failure is an engine rendering bug, not a converter bug.
The failure requires new schema fields.
The only new signal is a full supplemental corpus rerun with no bounded cluster, hypothesis, or classification.

Entry Points

/migrate-research
/migrate-research --top 20
/migrate-research --styles "apa,nature"
/migrate-research --resume

Setup (read first)

Before doing anything else, read:

docs/policies/STYLE_WORKFLOW_DECISION_RULES.md — failure classification and stop conditions
docs/guides/STYLE_WORKFLOW_EXECUTION.md — evidence ladder, cluster reduction, session loop

Autonomous Operation

Run the full research loop without pausing for approval. Only interrupt for Cargo.toml/Cargo.lock changes or git push origin main.

Workflow

Use the shared execution guide for the evidence ladder, cluster reduction, and stop conditions.
Keep the file focused on the converter-specific loop and migration session state.

Session State

Keep the existing lab/ scratch structure under .claude/skills/migrate-research/.

Output Contract

Report the chosen cluster, classification, before/after evidence, exact change made if any, and the final verdict on whether the pass should continue, stop, or escalate.

同仓库更多 Skills

同仓库

style-migrate-enhance

citum/citum-core

High-throughput migration waves converting priority parent CSL 1.0 styles to Citum with repeatable before/after metrics and migration-engine gap recommendations. Fidelity is the hard gate.

2026-07-0230

style-qa

citum/citum-core

Standardized QA gate for style work. Verifies fidelity, SQI drift, formatting defects, and regression surface. Produces approve/reject verdict with numbered findings.

2026-07-0230

style-tune

citum/citum-core

Iterative LLM hand-tuning loop for embedded-core styles. Drives a style to 100% fidelity and clean SQI. Both are hard gates. Seeded from migrate evidence, not from the converter output as a terminal deliverable.

2026-07-0230

migrate-research

citum/citum-core

Autonomous research loop for Citum migration fidelity gaps and converter-focused Rust fixes.

2026-07-0230

style-evolve

citum/citum-core

Single human-facing command for all Citum style work. Use whenever someone asks to fix, improve, convert, create, or tune a citation style. Routes to upgrade, migrate, create, or tune. Always use this rather than calling internal skills directly.

2026-06-2430

style-evolve

citum/citum-core

Public entrypoint for Citum style work. Activate on: 'upgrade', 'migrate', 'create', 'tune', any style authoring request, or any request to fix/improve/convert a Citum or CSL citation style. Route to the shared workflow docs and internal roles.

2026-06-2430

name	migrate-research
type	user-invocable, agent-invocable
description	Autonomous research loop that improves citum-migrate converter fidelity by iterating on Rust code in crates/citum_migrate/src/. Measures fidelity across a corpus of priority styles, clusters failure patterns, attempts targeted Rust fixes, validates with cargo nextest, keeps improvements, reverts failures. Does NOT modify style YAML.
model	sonnet

Migrate Research

Use This Skill When

Systematic converter improvement is the goal.
Multiple styles share the same migration failure pattern.
You want to close the gap between citum-migrate output and reference fidelity.
A style-fidelity follow-up needs bounded rich-input evidence to determine whether the next pass belongs in migration, style, processor, or adjudication.

Do Not Use When

Fixing a single style's YAML.
The failure is an engine rendering bug, not a converter bug.
The failure requires new schema fields.
The only new signal is a full supplemental corpus rerun with no bounded cluster, hypothesis, or classification.

Entry Points

/migrate-research
/migrate-research --top 20
/migrate-research --styles "apa,nature"
/migrate-research --resume

Setup (read first)

Before doing anything else, read:

docs/policies/STYLE_WORKFLOW_DECISION_RULES.md — failure classification and stop conditions
docs/guides/STYLE_WORKFLOW_EXECUTION.md — evidence ladder, cluster reduction, session loop

Autonomous Operation

Run the full research loop without pausing for approval. Only interrupt for Cargo.toml/Cargo.lock changes or git push origin main.

Workflow

Use the shared execution guide for the evidence ladder, cluster reduction, and stop conditions.
Keep the file focused on the converter-specific loop and migration session state.

Session State

Keep the existing lab/ scratch structure under .claude/skills/migrate-research/.

Output Contract

Report the chosen cluster, classification, before/after evidence, exact change made if any, and the final verdict on whether the pass should continue, stop, or escalate.