| name | research-report-reviewer |
| description | Use when a research draft or report exists and needs a pre-final review for unsupported claims, ambiguity, scope drift, or missing evidence before delivery. |
| metadata | {"short-description":"Findings-first review of a research draft"} |
OpenCode Runtime Notes
This skill is installed as an OpenCode-native SKILL.md. For runtime-backed
helpers, prefer the shared ai-agents-skills runtime root and the
AAS_RUNTIME_ROOT override instead of assuming a Codex-specific runtime
path.
Research Report Reviewer
Use this after a draft exists and before presenting research as final.
What to inspect
- the stated scope, question, exclusions, and any requested output format
- available structured artifacts such as
sources.jsonl, claims.jsonl,
guards.jsonl, delivery.json, source ledgers, analysis matrices, and report
evidence mappings
- unsupported or weakly supported claims
- missing dates or stale-time ambiguity
- scope drift relative to the original question
- places where observation and inference are blended together
- overconfident language that should be hedged or marked
incomplete analysis
- whether prior posts, templates, style guides, or supplied examples were
inspected before a format-matched draft
- whether the draft or workflow records an active writing-style profile from
writing-style-settings.md, plus math-manuscript-style.md when applicable,
including style_profile_ref, active_overlays, and
active_requirement_ids
- whether
style_applied: true is supported by workflow evidence rather than a
bare self-assertion
Output contract
Start with a visible section titled Review Findings.
Then give:
Verdict — BLOCK, FLAG, or PASS
Findings — the highest-signal issues first
Repairs — the minimum changes needed before delivery
Style — missing or inconsistent style_profile_ref, active_overlays,
active_requirement_ids, or style_applied records when relevant
If there are no issues, say so explicitly and keep the pass short.
Use references/reviewer-prompt.md as the detailed checklist.
Guardrails
- findings first, summary second
- focus on research quality, not copyediting
- prefer the smallest repair that makes the draft defensible
- if a gap cannot be closed, require explicit disclosure instead of pretending it is solved
- do not apply repairs, patch the reviewed report, create follow-on artifacts,
or continue into remediation after a review-only request unless the user
explicitly asks for those actions