메뉴
Produce the autoresearch report with full idea lineage — paper → claim → hypothesis → patch → metric change → status → memory update. Use to summarize an autoresearch run into an auditable, human-readable report at .autoresearch-vkf/session/report.md.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Run the continuous experiment loop, including idea tournament, best-first tree expansion, mandatory literature cadence, bold swings, automatic commits, and memory write-back.
Gather literature, extract structured claims, and verify their sources and codebase relevance as one trust-preserving pipeline.
Run an autonomous research or optimization loop with literature-grounded, verifiable long-term memory. Use for measurable optimization and for ranked research-plan ideation.
Generate grounded novel hypotheses from contradictions, complementary claims, and cross-domain mechanism transfers using the unified synthesize tool.
SOC 직업 분류 기준
| name | autoresearch-vkf-research-report |
| description | Produce the autoresearch report with full idea lineage — paper → claim → hypothesis → patch → metric change → status → memory update. Use to summarize an autoresearch run into an auditable, human-readable report at .autoresearch-vkf/session/report.md. |
Turn the run into an auditable story. The point of verifiable memory is that every result can be traced back to where the idea came from and forward to what was learned. Make that lineage explicit.
export_dashboard to write the browser views — progress.html (metric chart +
timeline) and dashboard.html (the idea-lineage graph) — and link them from the
report.research_status for the session experiments and memory lifecycle counts.recall_memory for the trusted claims, candidates, and negatives..autoresearch-vkf/memory/): paper, claim, and experiment
cards, and the transactions/ audit trail..autoresearch-vkf/session/report.mdInclude:
Goal & metric — objective, metric, direction, baseline → best achieved.
Headline result — the best outcome and the change that produced it.
Idea lineage — the heart of the report. For each idea that was tried, a chain:
Paper: <title> (<source_url>)
Claim: <claim:id> — <assertion> [mechanism: …]
Hypothesis: <intervention + prediction>
Patch: <what changed / commit>
Metric: <baseline> → <observed> (Δ …)
Status: win | loss | inconclusive
Memory: <claim belief x → y, state →>; experiment card <experiment:id>
What we learned — durable takeaways now in memory, including negative results ("X failed here because … — don't retry unless …").
Open directions — promising-but-untested ideas and gaps/contradictions worth exploring next run.
Provenance & trust — counts by lifecycle state; note anything flagged
stale by vkf freshness; confirm vkf validate passes at the bundle's
profile.
Honest over impressive. The differentiator of this system is that its memory is verifiable — so report what replicated, what didn't, and what is still just "the paper says so." Never inflate a 5-minute proxy win into a real win; say which it is.