ワンクリックで
quest-handoffs
Create durable CodexScientist project handoff documents, AGENTS.md files, researcher packages, and verified sync artifacts.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Create durable CodexScientist project handoff documents, AGENTS.md files, researcher packages, and verified sync artifacts.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | quest-handoffs |
| title | Research Quest Handoffs |
| description | Create durable CodexScientist project handoff documents, AGENTS.md files, researcher packages, and verified sync artifacts. |
| version | 1.0.1 |
| metadata | {"hermes":{"category":"research","tags":["handoff","codexscientist","research","agents","synchronization","verification"]}} |
| skill_role | companion |
Codex adapter note: this is a Codex-packaged CodexScientist support skill. Prefer public MCP
cs_*tools visible intools/list/cs_tool_schemafor durable root-bound research state, memory, artifacts, and recovery anchors; use admin CLI fallbacks only when explicitly following docs/ADMIN_CLI.md. Load it only when it is the relevant companion to the active stage. When a handoff becomes durable, record it withcs_artifact_recordas a milestone/report and include the absolute research root path in the payload.
Use this skill when the user asks for onboarding notes, a guide for another agent, an AGENTS.md, a quick-start document, or a synced handoff for a research/CodexScientist project.
Write the handoff into the quest filesystem. Do not leave it only in the final chat reply. Another agent should be able to start from the handoff file plus the referenced quest documents.
Create or update a quest-root AGENTS.md unless the project already has a clearer convention.
Include these sections when relevant:
For a single handoff file, verify with:
sha256sum source targetcmp -s source targetwc -l source targetstat -c '%n %s bytes %y' source targetFor full quest-copy verification, compare recursively by relative path and hash. Ignore .git/ internals such as .git/index, which can differ after status checks even when working content is equivalent.
CodexScientist/config/codex-native.yaml, unless the user explicitly asks.When a long-running quest's AGENTS.md, formal command document, or execution plan becomes too long or stale, prefer a single concise current-status entry (for example experiments/CURRENT_STATUS.md) plus short pointers from the long documents. Preserve source-of-truth plans and run evidence; do not keep appending chronological handoff blocks. See references/concise-current-status-maintenance.md.
When the user asks to send a quest to another researcher for analysis, create a compact no-code zip with a root README_FOR_RESEARCHER.md, manifest files, active idea docs, experiment plans, result reports, figures/tables, and small source-of-truth evidence artifacts. For external researcher analysis, default to a minimal tier (idea/protocol/status/report/final evidence only) and expand only on request; avoid bundling long command ledgers, full figure directories, or broad provenance if the user says the package is too cluttered. Exclude code, model weights/checkpoints, raw caches, PDFs unless requested, .cs/, and secondary workspace copies under current no-sync constraints. See references/no-code-researcher-handoff-zip.md.
When the user asks for all necessary information in one Markdown file for another researcher to analyze from scratch, write a self-contained quest-root handoff such as RESEARCHER_ZERO_START_ANALYSIS.md. Include the project question, method route, current experiment state, key results, safe/unsafe claims, limitations, and suggested analysis order. If the recipient will not receive other files, do not lean on evidence paths; embed the judgment-critical method facts directly: key function formulas/calculations, active-vs-normative implementation differences, candidate enumeration, data/batching, trainable/frozen parameters, optimizer/schedule/eval cadence, verification fields, and explicit caveats about what is not implemented or not run. Read source-of-truth docs, implementation specs/code, configs, and final evidence while drafting, then loop back to patch omissions before validating. See references/single-file-zero-start-researcher-handoff.md.
See references/codexscientist-quest-handoff-sync.md for the session-derived checklist and verification pattern.
Write CodexScientist-aware implementation, experiment-roadmap, code-only, and formal-command plans with TDD and durable quest outputs.
Use when a quest needs one or more follow-up runs such as ablations, robustness checks, error analysis, or failure analysis after a main experiment.
Use when a quest needs to attach, import, reproduce, repair, verify, compare, or publish a baseline and its metrics.
Use when the quest needs an explicit go, stop, branch, reuse-baseline, write, finalize, reset, or user-decision transition with reasons and evidence.
Execute CodexScientist project experiment command documents with durable cs_* logging, manifest validation, and cautious baseline-gate handling.
Use when a quest is ready for a concrete implementation pass or a main experiment run tied to a selected idea and an accepted baseline.