research-gpt
25Agent
Theory-focused research agent for theoretical foundations, concept definitions, proof validity, and fundamental limitations. Triggers: explain theory behind, what is X, prior work on, theoretical analysis, concept definition.
Agent
Theory-focused research agent for theoretical foundations, concept definitions, proof validity, and fundamental limitations. Triggers: explain theory behind, what is X, prior work on, theoretical analysis, concept definition.
Agent
Multi-perspective quality verification with explicit rubrics. Coordinates independent critic agents (Security, Performance, Correctness, Maintainability) to avoid self-bias. Triggers: validate against criteria, multi-perspective check, rubric evaluation, quality verification. SRP: Rubric-based validation only (no…
Agent
Cross-model idea validation for methodology and theory correctness. Coordinates research-gpt and research-gemini for multi-perspective verification. Triggers: validate idea, is this approach correct, methodology check, theory validation.
Skill Claude CodeCodex
Pre-merge/commit code review for correctness, regressions, policy compliance, and maintainability. Triggers: review code, code review, review before merge, check this PR, review changes.
Skill Claude CodeCodex
Strict workflow for Git commits: inspect changes since last commit, generate a Conventional Commits message, show summary, and require explicit user confirmation before commit.
Skill Claude CodeCodex
Experiment result analysis with reproducible metrics, charts, and Korean decision summaries. Triggers: analyze results, compare metrics, visualize data, experiment analysis.
Skill Claude CodeCodex
Multi-source, citation-backed research with Korean reports and traceable evidence. Triggers: research topic, investigate, deep research, multi-source analysis.
Skill Claude CodeCodex
Mediator-first deep reasoning protocol for high-difficulty tasks. Triggers: deep-think, multi-hypothesis reasoning, structured refinement, mediator triage.
Skill Claude CodeCodex
Create and maintain project documents (drafts, final, reference) with strict language and path policy. Triggers: write doc, create documentation, document this, publish report.
Skill Claude CodeCodex
Mediator-first multi-perspective council protocol for disagreement surfacing and synthesis. Triggers: dt-council, model council, diverse perspectives, mediator redirect.
Skill Claude CodeCodex
Generate local skills from external sources using quarantine-first security workflow. Triggers: import external skill, generate skill from docs, external documentation to skill.
Skill Claude CodeCodex
Monthly prompt engineering paper catalog update with scoring rules and safe add/retire operations. Triggers: update paper catalog, refresh paper DB, paper database update.
Skill Claude CodeCodex
Create or refine single-responsibility SKILL.md files with valid frontmatter and executable protocol. Triggers: create new skill, new SKILL.md, improve skill, skill authoring.