← ClaudeAtlas

research-doclisted

Write a self-contained, template-conformant research document with Chain-of-Verification grounding, adversarial review, live citation validation, a provenance ledger, and claim-tier tags. Use for a scoped "research X and write it up" request, inline or through an installation-provided research adapter.
sergeiwallace/skills · ★ 0 · AI & Automation · score 73
Install: claude install-skill sergeiwallace/skills
# research-doc Write a research document using the bundled prompt template and document stub. The methodology prioritizes factual grounding over coverage: adversarial critics, live citation validation, claim-to-quote-to-verdict provenance, Chain-of-Verification (CoVe), and explicit claim tiers. ## Invocation and optional integrations **Usage:** `/research-doc <topic | path-to-research-doc> [--model <model-preference>] [--inline] [--refine]` - `--inline` writes the document in the driving session. It is always available and is the default when no delegation adapter is configured. - `--model` records a model preference for the current environment or its optional delegation adapter. Use only a value supported by that installation. - `--refine` updates an existing document. For a bare topic, first look for a substantively related document in the project locations available to the current installation; refine it instead of creating a near-duplicate when one exists. Before authoring, this bundled resolver can validate the documented options and emit their JSON envelope. It performs no research, routing, or file creation: ```sh uv run --script "${CLAUDE_SKILL_DIR}/scripts/research_doc_resolve.py" \ "<topic-or-path>" [--model <model-preference>] [--inline] [--refine] ``` ### Optional tracking adapter Tracking is optional and never blocks research. To connect an installation's tracker, set `RESEARCH_DOC_TRACKING_ADAPTER` to the absolute path of a project-owned execut