Skip to content

Ar9av/PaperOrchestra / skills/paper-orchestra

paper-orchestra

Verified

Orchestrate the full PaperOrchestra (Song et al., 2026, arXiv:2604.05018) five-agent pipeline to turn unstructured research materials (idea, experimental log, LaTeX template, conference guidelines, optional figures) into a submission-ready LaTeX manuscript and compiled PDF. TRIGGER when the user asks to "write a paper from my experiments", "turn this idea and these results into a paper", "generate a conference submission", "run paper-orchestra on X", or otherwise wants the end-to-end paper-writing pipeline. Coordinates the outline-agent, plotting-agent, literature-review-agent, section-writing-agent, and content-refinement-agent skills.

references/outline-reconciliation.md

View on GitHub
# Outline Reconciliation (Step 3.5)

**Inspired by:** AutoSci (arXiv:2605.31468) — hypothesis refinement based on early findings.

## Why this step exists

The outline is generated in Step 1 from `idea.md` and `experimental_log.md` alone — before
any literature is actually found. Step 3 (Literature Review) may discover that:

- A Related Work cluster the outline assumed was well-populated returns few or no verified papers
- The baseline comparisons framed in `section_plan` rely on papers that failed S2/Crossref
  verification
- The actual landscape of citations warrants splitting, merging, or re-ordering clusters
- The Introduction strategy should emphasise different foundational context based on what
  was found

Without reconciliation the section-writing agent receives an outline whose `section_plan`
may contradict the actual `citation_pool.json`. The reconciliation step closes this gap with
**one lightweight LLM call** after Step 3 completes, before Step 4 begins.

## What gets updated

Only `section_plan` is updated — **never** `plotting_plan` (Step 2 may still be running)
and **never** `intro_related_work_plan` (Step 3 has already acted on it).

Permitted changes:
- Reframe section claims to match what the verified citation pool actually supports
- Adjust or remove baseline comparisons that cannot be backed by found papers
- Reorder or rename subsections to reflect the actual literature structure
- Add a note to a section that a certain comparison is "concurrent work" rather than
  a beaten baseline (per the TIMELINE RULE)

Forbidden changes:
- Adding or removing top-level sections defined in `template.tex`
- Changing any field in `plotting_plan`
- Changing any field in `intro_related_work_plan`
- Inventing new citation hints for papers not in `citation_pool.json`

## Input / Output

| File | Role |
|---|---|
| `workspace/outline.json` | Read — original outline (Step 1 output) |
| `workspace/citation_pool.json` | Read — what was actually verified (Step 3 output) |
| `workspace/cross_verification_report.json` | Read — confidence tiers (Step 3 output) |
| `workspace/drafts/intro_relwork.tex` | Read — what claims Step 3 actually wrote |
| `workspace/outline_reconciled.json` | Write — updated outline for Step 4 |

If `citation_pool.json` is absent (Step 3 is still running or failed), skip this step
and use `outline.json` directly in Step 4.

## Prompt for the reconciliation call

Load `skills/outline-agent/references/prompt.md` as the base system prompt, then prepend:

```
RECONCILIATION MODE — do not regenerate the full outline.

You are reviewing the original outline.json against what was actually found
during the literature review. Your only task is to update the `section_plan`
array so that every claim in each section is backed by the verified citation
pool.

RULES:
1. Output a complete JSON object with the same three top-level keys as outline.json.
   Copy `plotting_plan` and `intro_related_work_plan` VERBATIM — do not change a
   single character.
2. Update only `section_plan` entries where the original content_bullets reference
   papers or comparisons that (a) failed verification, or (b) are post-cutoff
   (concurrent work only).
3. For each changed bullet: keep the scientific claim, adjust only the framing
   (e.g. "outperforms [X]" → "compares favourably with the concurrent work [X]",
   or drop the comparison if [X] is entirely absent from citation_pool.json).
4. Do not add subsections or remove top-level sections.
5. Output raw JSON only — no prose, no markdown fences.
```

Input context to provide:
- `outline.json` (full)
- `citation_pool.json` (paper titles + bibtex_keys only — no full abstracts needed)
- `cross_verification_report.json` (only the `low` and `conflict` tier entries)

## Validation

After the call, validate the output:

```bash
python skills/outline-agent/scripts/validate_outline.py workspace/outline_reconciled.json
```

If validation fails, fall back to `outline.json` for Step 4 and log a warning.

Diff the two files to produce a human-readable reconciliation summary:

```bash
python skills/paper-orchestra/scripts/diff_outlines.py \
    --original  workspace/outline.json \
    --reconciled workspace/outline_reconciled.json \
    --summary   workspace/reconciliation_summary.md
```

Report the summary to the user before starting Step 4.

## When to skip

- Step 3 failed or produced an empty citation pool → skip, use `outline.json`
- `cross_verification_report.json` shows zero `low`/`conflict` entries AND
  all section_plan citation hints are already present in `citation_pool.json`
  → skip (nothing to reconcile), use `outline.json`
- Host does not support a 4th parallel call at this stage → skip, use `outline.json`

Skipping is safe. Reconciliation is a quality improvement, not a hard gate.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall