p-aldighieri

Fan out a 4-member council (1 Codex thinking-high + 1 Gemini Pro + 1 Claude Opus + 1 ChatGPT Extended Pro) on a stalled re-attack. Preserves all four independent memos; hands off to the regular Strategy Searcher for pure selection. Opt-in (attempt ≥2), ~3× the cost of a single search.

Harvest a completed ChatGPT chat's last assistant message into a response file. Use when a role finished but the response file wasn't captured (e.g., orchestrator lost the chat URL, or the submit script died after the model responded).

Create or update {PROOF_REPO}/referee_targets.yaml — the target-journal registry the paper_referee role consults to judge whether the result clears the bar at the journals you might submit to. Optional; without it the referee falls back to a generic publishability check.

Use when working in MathPipeProver on ChatGPT project preparation, durable source refresh, project-source inspection, or browser-backed external_agent submission. Prefer scripts/chatgpt_browser_agent.sh prepare/submit over ad hoc UI work, enforce the Extended Pro target, and keep durable sources separate from temporary attachments.

Read-only one-shot inspection of a live ChatGPT chat — returns generating flag, assistant turn count, last assistant message length and tail. Use during long-running roles to check whether the chat is still working or done.

Verify or set the ChatGPT composer to the Extended Pro target (reasoning Pro + model 5.5; pill reads "Extended Pro" or "Pro"). Required before any analytical submission.

Add or remove durable files in a ChatGPT project's Sources tab. Use after a verified result lands, after a route pivot, or when the durable proof-state file has been updated locally.

Submit a proof role prompt to a ChatGPT project via CDP and monitor for completion. Verifies Extended Pro (or toggles Deep Research for the literature role), sends, reports the chat URL, and provides recovery guidance for long-running roles.

End-to-end first-draft orchestrator. Takes an **external target folder path** (the paper lives outside this repo). Sets up missing scaffold pieces, then runs the role agents in two phases: Phase 1 (parallel) writes related-lit and model-prose. Phase 2 (sequential) writes results-prose, then optional discussion, then intro (incl. closing if warranted), then abstract last. Pauses for review between phases. Use when the user says "draft <path>", "draft `~/Documents/papers/foo`", or similar.

Verify EVERY entry in a paper's `refs.bib` against Crossref's canonical record. Catches wrong DOIs, years, volumes, issues, pages, venues, and author lists; flags missing DOIs with the record to add; and detects when a "working paper" now has a published version. **Report-only — never edits the .bib.** **Does not check math or prose.** Use before any "final" compile / submission, or whenever the user says "check the citations", "verify the bib", "crossref the references", "are the citations right".

Compile a paper in `<target>/tex/` to PDF. Runs pdflatex → bibtex → pdflatex → pdflatex, parses the .log for errors, surfaces a focused error report. Use whenever the user says "compile", "build the PDF", or after any non-trivial edit to the .tex sources.

Bootstrap a new paper. Asks for an external target path, creates the directory, copies the scaffold, and walks the user through filling in the brief. Use when the user says "new paper", "start a paper", "I want to write a paper", "set up a new paper", or when they're new to the repo and need onboarding. This is the **first-run entry point**.

Audit the current draft for craft and integrity violations. Runs `integrity-gatekeeper` and `exposition-gatekeeper` agents in parallel, synthesizes findings into Critical / Important / Minor. **Reports only — never auto-fixes.** **Does not check math.** Use after any substantive draft pass.

Tighten the "why does this matter" thread running through abstract + intro P1-P2. Targets motivation density, not prose polish. Use after the first intro draft if motivation is weak.

Purge pipeline-internal references from the user's pre-baked proofs in `<target>/proofs/`. Outputs a `_cleaned/` directory + a diff report. **Never touches the math itself** — only removes meta-prose referring to the proof-checking pipeline (e.g., "in iteration 3 of the verification", "as shown in the previous step of the pipeline"). The user reviews the diff and decides whether to swap in the cleaned files.

Draft or revise the Related Literature section. Uses a verified citation universe (`refs.bib` + `brief.md` anchors + `references/pdfs/`). **Never invents citations.** Marks unresolved claims with `[CITATION NEEDED]`. Use when the user says "write the related lit", "do a lit review pass", or after the main results are stable.