메뉴
Alias for the AnchoredAutoResearch Codex planning wizard. Equivalent intent to $research-plan. Supports --effort low|middle|high and --search N for literature-pool size.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
AnchoredAutoResearch planning wizard for Codex. Builds the PLAN section with scope, metrics, verify/guard commands, predictions, and human approval gates. Supports --effort low|middle|high and --search N for literature-pool size.
AnchoredAutoResearch: Phase-gated research protocol with human checkpoints, prediction discipline, knowledge sources, and autonomous execution. Manages the full research cycle: plan → execute → interpret.
AnchoredAutoResearch adopt skill for Codex. Anchors the protocol to an in-progress project (existing LaTeX + git + logs). Imports state as read-only Cycle 1; real cycles start at Cycle 2.
Alias for the AnchoredAutoResearch Codex adopt skill. Equivalent intent to $research-adopt.
Alias for the AnchoredAutoResearch Codex paper skill. Equivalent intent to $research-paper.
AnchoredAutoResearch paper skill for Codex. Writes, edits, and compiles a venue-specific LaTeX paper from the Research State Document. Strict side-effect command — never touches RSD.md, never advances phase.
| name | research:plan |
| description | Alias for the AnchoredAutoResearch Codex planning wizard. Equivalent intent to $research-plan. Supports --effort low|middle|high and --search N for literature-pool size. |
| version | 1.0.0 |
Use this skill when the user invokes $research:plan in Codex.
Optional flags:
--effort low|middle|high controls ideation breadth. Default is low
(current behavior). See ### Effort levels in
../research/references/phase-protocol.md for the per-level protocol.
Reject any other value with: "unknown effort level — use low, middle,
or high".--search N sets the literature discover-pool size when this
invocation triggers /research:context search <topic>. Integer in
[10, 500]; clamped. Default by effort: low ignores it, middle
honors if passed, high defaults to N=200. Deep-read count is
min(30, N // 5). See knowledge-sources.md "Search mode" for the
two-tier protocol. Reject non-integer values with: "--search must
be a positive integer in 10..500".Follow the same workflow as the research-plan skill:
AGENTS.mdRSD.mdcontext/SOURCES.md if it exists--effort and --search flags if present and branch on them
per the ### Effort levels section in
../research/references/phase-protocol.md (LOW ignores --search,
MIDDLE honors if passed, HIGH defaults to --search 200)../research/references/phase-protocol.md../research/references/knowledge-sources.mdEffort level: and
Alternatives considered: fields), compile, commit, and stop for human
approval