使用请引用 / Please cite: Haocheng Wang (2026), Communication Research Skills for Codex. https://github.com/haochengw372-hash/communication-research-skills. Machine-readable citation: CITATION.cff. Citation is requested as an academic courtesy, not an additional license condition.

Communication Research Skills for Codex

A maintainable Codex Skill suite for communication research and computational communication studies (计算传播学). It differs from general CS or social-science packages by routing research through a communication reasoning graph — phenomenon, literature, theory, construct, mechanism, RQ/H, operationalization, measurement, design, evidence, and theoretical contribution — and by enforcing seven domain gates.

What is inside

Eighteen installable Skills live under skills/, one directory per Skill:

Research flow

flowchart LR
    P["Phenomenon"] --> L["Literature"]
    L --> T["Theory"]
    T --> C["Construct"]
    C --> M["Mechanism"]
    M --> R["RQ / Hypotheses"]
    R --> O["Operationalization"]
    O --> S["Measurement"]
    S --> D["Design / Identification"]
    D --> E["Evidence"]
    E --> G["Theoretical contribution"]

Seven gates sit on this graph: Theory, Construct, Measurement, Design/Identification, Novelty, Contribution, and Evidence-to-Claim. Each is owned by a specialist Skill and enforced by the orchestrator before the next transition.

Installation

Copy the whole suite into your Codex skills directory:

scripts/install.sh                 # installs to ~/.codex/skills
scripts/install.sh --dest /tmp/x   # install to a test destination
scripts/uninstall.sh               # removes the eighteen installed folders

Manual installation is equivalent: copy each folder in skills/ into ~/.codex/skills/<folder-name>/. After installation, refresh or restart Codex so the new Skills appear in the catalog.

The orchestrator lives at skills/communication-research-workflow/ and the domain Skills in sibling folders under skills/.

Quick start

Invoke the orchestrator with a realistic research request:

$communication-research-workflow

Plan only. Topic: compare how responsibility for an environmental risk is framed
across two digital platforms over time. Classify the research goal before choosing a
method, propose the lifecycle, route the specialist Skills, and list the evidence and
measurement gates that must pass before analysis.

For a narrow capability, invoke the domain Skill directly, for example $communication-theory, $communication-content-analysis, $communication-network-analysis, $communication-causal-inference, or $communication-writing; use $scholarly-access with a DOI list.

Project-control layer

Long-running projects can be governed without moving existing files. The orchestrator creates the same non-invasive .codex-research/ layer used by its upstream:

research-project/
├── existing data, code, manuscripts, and outputs
└── .codex-research/
    ├── project.yaml
    ├── state.yaml
    ├── windows.yaml
    ├── decisions.md
    ├── contracts/
    ├── handoffs/
    ├── audits/
    └── snapshots/

Only the controller writes authoritative governance state. Executors and auditors return immutable handoffs with hashes and fresh verification.

Personalization and case study

The generic Skills never embed personal research defaults. Private or project-specific preferences belong in a personal derivative or project AGENTS.md, following references/personalization-guide.md. A full worked case study of the role and handoff lifecycle is in references/complete-case-study.md.

Unified methodology, CSS layer, and demo

• METHODOLOGY.md is the single methodology the suite operationalizes: the reasoning graph, the seven gates, goal-first method routing, evidence-to-claim discipline, the five-layer scholarly-access policy, and the end-to-end workflow.
• Computational social science sits under the communication layer. The general CSS theory anchors are in skills/communication-theory/references/css-theory.md, and the general CSS method families are in skills/communication-method-router/references/css-methods.md.
• demo/RESEARCH_DEMO.md contains three synthetic, replaceable examples covering computational content analysis, network/diffusion analysis, and communication experiments. The generated plans demonstrate archetype-level window graphs without encoding a preferred topic or project. Network, causal, temporal, multimodal, spatial, and simulation analyses each have a dedicated specialist Skill.

License

Original contributions by Haocheng Wang are licensed under Apache-2.0. See LICENSE and NOTICE.md for the retained upstream MIT notice, attribution, and methodological inspiration.