add-paper

name	add-paper
description	Triage skill — routes academic-style papers to `paper-processing-academic`, institutional-style reports to `paper-processing-institutional`, non-paper documents to `document-processing` (deferred). Step 0 triage runs here; the routed per-protocol skill handles Steps 1-6.
nori_researcher_source	nori-skillsets add-paper v1.0.0 (ported to claude_researcher in 0bbd419, 2026-05-10)
aitaxbid_source	~/code/AITaxBID/skills/paper_processing.md@e0a736d (2026-05-02)

Runtime detection

Before following the rest of this skill, determine your environment:

if [ "$IS_SANDBOX" = "yes" ] || [ -d "/mnt/skills/public" ]; then
  echo "claude.ai sandbox"
elif [ "$CLAUDECODE" = "1" ]; then
  echo "Claude Code"
else
  echo "unknown — surface to user before proceeding"
fi

Both environments set positive markers; the probe checks for either side affirmatively rather than inferring from absence. If neither fires, something is misconfigured (env vars stripped, custom shell, etc.) and silently picking a branch is worse than surfacing the question.

If claude.ai sandbox: the user's project repo is already cloned at /home/claude/<REPO>/ per RESEARCHER.md §2.0b — run the git add / git commit / git push commands in this skill directly from that working tree. Translate local skill paths like /Users/<user>/.claude/skills/... to the template clone at /home/claude/.claude_researcher_template/template/skills/.... Only if the §2.0b clone failed (degraded REST fallback, surfaced to the user) do you translate git add / git commit / git push into the Contents API recipes from your Project Instructions.

If Claude Code: follow the skill body as-is.

If unknown: stop and surface to the user. Don't guess which environment you're in — the cost of a wrong guess (operating against the wrong working tree, or using the wrong write path for the environment) is higher than the cost of one round-trip clarification.

*CRITICAL* Add the following steps to your Todo list using TodoWrite:

0. Triage — Protocol A or Protocol B (Step 0)
1. Load and follow the routed per-protocol skill (Dispatch at end of Step 0)

Scope

This skill is a triage dispatcher. It runs Step 0, then routes to one of three target skills that do the actual work:

• Academic-style papers — research-oriented documents with original empirical or theoretical contribution (journal articles, working papers, dissertations, white papers structured as research). Routed to paper-processing-academic by Step 0 triage.
• Institutional-style reports — substantive analytical documents that synthesize evidence, position a framework, or advise on policy, but do not present the authors' own original research with a stated hypothesis (G20 background notes, IMF/World Bank/OECD/UN flagship reports, multilateral working-group papers, regional development bank policy reports). Routed to paper-processing-institutional by Step 0 triage.
• Legislation, government regulatory documents, terms of reference, and consultant deliverables tied to a single project — routed to document-processing. (Note: document-processing is deferred per Plan 02 Wave 5; the pointer is currently aspirational — if you hit this branch, fall back to manual handling and surface to the user.)

Where config keys live. The four PROJECT_QUESTION / CONDITIONAL_SECTION / BIB_FILE / PAPERS_INDEX keys live in the research repo's STATUS.md under ## Project parameters. The two paper_naming keys (paper_naming.academic_format, paper_naming.institutional_format) live in the user's personal_info.md under "Operating preferences". The paper_summaries.structure knob also lives in STATUS.md project parameters. The per-protocol skills each reference back to this section for the full schema.

Adding a Paper

Announce at start: "I'm using the Add Paper skill — first I'll triage the document, then dispatch to the appropriate per-protocol skill."

Step 0: Triage — academic-style or institutional-style?

Open the document and answer three quick questions:

1. Does it have an abstract (vs. an executive summary)?
2. Does it pose a research question or hypothesis?
3. Does it report new estimates the authors produced from data they analyzed (rather than synthesizing others' findings)?

• Two or more "yes" → Protocol A (academic-style).
• Two or more "no" → Protocol B (institutional-style).

For borderline cases, exercise judgment and flag the call to the user in the conversation. Common borderline cases:

• Multilateral working papers that have abstracts and methods sections but lean heavily on policy framing — usually Protocol A.
• Institutional monographs with chapter-level empirical work — usually Protocol B at the document level (the value is the synthesis); if a chapter is heavily reused on its own, consider creating a separate Protocol A summary for that chapter and cross-linking.
• Country case studies with descriptive but not causal analysis — usually Protocol B.
• IDB Discussion Papers, Technical Notes, and Working Papers — most have research-paper structure and are Protocol A; monographs and synthesis pieces are Protocol B. Triage on structure, not on the publisher.
• Reports without named individual authors (e.g., OECD secretariat reports, UN agency reports) are unambiguously Protocol B.
• Reports with named authors but commissioned by an institution (e.g., "Prepared by Smith and Jones for the G20") — the cover usually names the institution as the publisher and the individuals as the production team. Treat as Protocol B.

The pipeline shape is identical across protocols. What differs:

Step	Protocol A	Protocol B
Filename	paper_naming.academic_format	paper_naming.institutional_format
Summary section (a)	Thesis, research question, contribution	Purpose, commissioning context, position
Summary section (b)	Data, sample, identification, effect sizes	Document type, frameworks/databases, headline findings, policy framework
Summary section (d)	Standard "relevance" framing	Same plus "what position does this represent" + cross-references
BibTeX entry type	@article / @unpublished	@techreport / @book / @inbook

For legislation, government regulatory documents, terms of reference, and consultant deliverables, route to document-processing instead — not the paper-processing skills.

Dispatch:

• Protocol A → academic → Read template/skills/paper-processing-academic/SKILL.md and follow it from Step 1 onward. Step 0 (this skill) is already complete.
• Protocol B → institutional → Read template/skills/paper-processing-institutional/SKILL.md and follow it from Step 1 onward. Step 0 (this skill) is already complete.
• Non-paper document (legislation, regulatory docs, ToRs, consultant deliverables) → use document-processing (currently deferred per Plan 02 Wave 5; if you hit this branch, fall back to manual handling and surface to the user).

Common Mistakes

Forgetting Step 0 triage

• Problem: Skipping triage routes the document through the wrong protocol — wrong filename convention, wrong section emphasis in the summary, wrong BibTeX entry type.
• Fix: Always start with Step 0. If borderline, flag the call to the user before proceeding rather than silently picking a branch.

Trying to do the full workflow in add-paper

• Problem: After Step 0, the agent ignores the Dispatch instruction and tries to execute Steps 1-6 from this file. Those steps no longer live here — they're in the per-protocol skills.
• Fix: After Step 0, read and follow the routed per-protocol skill (paper-processing-academic or paper-processing-institutional). This skill is intentionally thin; the real work is in the routed file.

Reading config keys from the wrong file

• Problem: Schema is split per Tier C — user-level paper_naming.* lives in personal_info.md; per-project PROJECT_QUESTION / CONDITIONAL_SECTION / BIB_FILE / PAPERS_INDEX / paper_summaries.structure live in STATUS.md ## Project parameters. The per-protocol skills each point back to this Scope section rather than duplicating the schema.
• Fix: Read user-level keys from personal_info.md; read per-project keys from the research repo's STATUS.md.