Academic Pipeline is a lightweight orchestrator for a ten-stage academic workflow. It coordinates the deep-research, academic-paper and academic-paper-reviewer skills, detects which stage you are at, recommends a mode and tracks state between stages.
The stages run research, writing, an integrity check, review, revision, re-review, re-revision, a final integrity check and finalisation. Integrity checks are mandatory, and the skill produces audit artifacts so each step can be traced.
What You Get
- A stage-by-stage workflow from research to a finalised manuscript
- Mandatory, coverage-bounded integrity checks before review and before finalising
- Two-stage peer review with revision and re-revision in between
- Auditable quality-assurance artifacts and a recorded state you can resume from
Common Use Cases
- Taking a paper from research question to submission-ready draft in one guided flow
- Resuming a long manuscript project in a new session from its recorded stage
- Enforcing integrity checks and two review rounds on a team’s papers
- Coordinating research, writing and review skills without choosing each mode by hand
What Makes It Unique
It is the conductor for the other skills in the suite and keeps no substantive logic of its own. Its value is the control around the work: mandatory integrity checkpoints, two review rounds, an optional context reset at checkpoints through a recorded passport hash, and an optional audit-artifact gate switched on by environment variables. A routing step asks for clarification when your materials span several stages.
Under The Hood
The skill loads a very long SKILL.md that holds stage rules and routing, plus agent files, references and templates for each checkpoint. It dispatches the research, writing and review skills and writes stage outputs, state records and audit artifacts as files in your project. Two opt-in environment variables switch on context-reset boundaries and the audit artifact gate. The other skills it calls search literature and check citations, so a full run reaches external sources through them. Pasted and fetched text is treated as data.