How yha9806/academic-writing-toolkit works
AWT is a local-first toolkit for evidence-controlled academic writing, operated through agent skills and deterministic local checks. The principal workflow moves from reading and noting sources, through mapping and integrating evidence into chapters, to auditing claims and references, then author review and export. The scaffold creates a thesis workspace; audit scripts provide internal validation. README-described skills are included, but their implementation is not present in the supplied tree. The experimental writing loop and archived governance tools are excluded because they are not part of the current main product.
Main entry points: skills Audit manuscript · awt.mjs Workspace scaffold · audit-citations.py Citation audit · audit-logic.py Logic audit
Components 19
Writing workflow
- Read and note
.claude/skills - Reading notes
- Map sources
.claude/skills - Integrate evidence
.claude/skills - Chapter drafts
- Clean-room review
.claude/skills - Export
.claude/skills - Word and ZIP
Evidence assurance
- Audit manuscript
.claude/skills - Citation audit
scripts/audit-citations.py - Logic audit
scripts/audit-logic.py - Content audits
scripts/audit-public-content.py - English audit
scripts/audit-british-english.py - Verify references
.claude/skills
Workspace setup
- Workspace scaffold
scaffold/awt.mjs - Thesis workspace
Other components
- Author
- Agent host
- Online metadata
Connections 25
- Author → Agent host: directs work
- Agent host → Read and note: runs skill
- Read and note → Reading notes: writes notes
- Agent host → Map sources: runs skill
- Map sources → Reading notes: reads notes
- Map sources → Chapter drafts: maps coverage
- Agent host → Integrate evidence: runs skill
- Integrate evidence → Reading notes: reads evidence
- Integrate evidence → Chapter drafts: updates drafts
- Agent host → Audit manuscript: runs checks
- Audit manuscript → Citation audit: runs audit
- Audit manuscript → Logic audit: runs audit
- Audit manuscript → Content audits: runs audit
- Audit manuscript → English audit: runs audit
- Audit manuscript → Chapter drafts: reads draft
- Audit manuscript → Reading notes: checks evidence
- Agent host → Verify references: runs skill
- Verify references → Online metadata: checks metadata
- Agent host → Clean-room review: runs skill
- Clean-room review → Chapter drafts: reviews draft
- Author → Clean-room review: checks findings
- Agent host → Export: runs skill
- Export → Word and ZIP: produces files
- Author → Workspace scaffold: requests setup
- Workspace scaffold → Thesis workspace: creates workspace