Full Research Pipeline (Humanities & Social Sciences)

End-to-end research workflow for: $ARGUMENTS

This is the non-interactive sibling of /prism-pipeline. It runs the same discipline-aware stages but omits the Soul Protocol (soul interview, soul-enriched gates, soul audit). Use /prism-pipeline when you want the human-checkpoint and authorial-voice layer.

Constants

• DISCIPLINE = auto — When auto, Stage 0 runs /prism-pipeline — mode: detect-only to detect from input and project context and write PRISM_PROFILE.md. Override with an explicit ID: law, sports-law, economics, social-sciences, humanities, philosophy, logic, education, management, communications. See shared-references/prism-routing.md.
• AUTO_PROCEED = false — When false, pause at each stage boundary for confirmation. When true, auto-select and continue. If detection fails and AUTO_PROCEED=true, default to humanities.
• AUTO_WRITE = false — When false, stop after critical review and leave paper writing to an explicit /paper-writing call.
• QUALITY_GATES = standard — off skips evidence/citation gates; standard treats failures as advisory; strict blocks paper writing until critical failures are fixed.

Pipeline Overview

PRISM detect → /research-lit → research questions → develop evidence → /research-review → /paper-writing (optional)

Stage 0: Discipline Detection

Run /prism-pipeline "$ARGUMENTS" — mode: detect-only (or honor an explicit DISCIPLINE) to produce PRISM_PROFILE.md. All downstream skills read it for discipline-specific routing.

Stage 1: Literature Review

/research-lit "$ARGUMENTS"

/research-lit reads PRISM_PROFILE.md and adapts source priority, venue filtering, and terminology. Humanities and law weight foundational works and primary sources (statutes, cases, canonical texts) over recent preprints.

Stage 2: Research-Question Formulation

Formulate 3–6 candidate research questions / theses grounded in the Stage 1 literature, then validate:

/novelty-check [candidate questions]
/research-review [top candidates]

If AUTO_PROCEED=false, present the ranked candidates and let the user choose before continuing.

Stage 3: Develop the Evidence Base

Build the evidence appropriate to the discipline (read PRISM_PROFILE.md):

• Law / sports law: legal issue map; statutes/cases/rules; comparative matrix
• Philosophy / logic: argument reconstruction; formalized inferences; objection/response map; for formal work, proof obligations
• Economics / social sciences / education / management: dataset assembly, empirical/identification strategy, analysis write-up; or survey/interview/coding protocols
• Humanities / communications: source corpus, analytical framework, close reading

Document findings in RESEARCH_LOG.md.

Stage 3.5: Evidence & Citation Gates

Run unless QUALITY_GATES=off (advisory under standard, blocking under strict):

• Law / sports law: statute-case-mapper → comparative-law → argument-stress-test → citation-audit (L3)
• Philosophy / logic (formal): /proof-checker, /formula-derivation
• Philosophy / logic (argumentative) & humanities: argument-stress-test
• Empirical social science / economics / education / management: /paper-claim-audit
• All disciplines: /citation-audit before submission if a bibliography exists

Stage 4: Critical Review

/research-review "$ARGUMENTS — [chosen thesis]"

The reviewer uses the discipline's review dimensions from PRISM_PROFILE.md. Hand specific claims to /argument-stress-test for a deeper adversarial pass. Output: review-stage/RESEARCH_REVIEW.md.

Stage 5: Paper Writing (optional)

Skip unless AUTO_WRITE=true or the user requests it.

/paper-writing "report-or-research-log"

Internally: /paper-plan → /paper-write → /paper-compile, with /research-refine for iterative polish after the evidence/citation gate is clean.

Side Tracks

• Funding: /grant-proposal branches from a validated research question; parallel to the publish track.
• Submission response: /rebuttal consumes external reviews; never invent new evidence.
• Presentation: /paper-slides, /paper-poster consume a compiled, audited paper.
• Collaboration: /overleaf-sync is a transport layer only.

Key Rules

• Profile-driven: every stage reads PRISM_PROFILE.md; if missing, fall back to humanities defaults.
• No fabricated evidence: claims must trace to grounded sources, cases, data, or valid arguments.
• Specialist skills are optional: if a named slash skill is unavailable, search local skill roots (.claude/skills, .codex/skills, .agents/skills, Prism-H/skills) for */SKILL.md and execute the closest equivalent.
• For the human-checkpoint and authorial-voice layer, use /prism-pipeline instead.