Claude handoff prompt

Copy everything below into a fresh Claude session running in this workspace.

I want you to rerun the Braskem restructuring as a rigorous multi-agent negotiation using public information through August 15, 2026 and the research already saved in this workspace.

Required preparation

First read completely:

1. .context/agent-memory/README.md
2. .context/agent-memory/petrobras.md
3. .context/agent-memory/braskem-management.md
4. .context/agent-memory/ig4.md
5. .context/agent-memory/creditors.md
6. .context/agent-memory/brazil-political-context.md

The underlying local evidence is in:

• .context/june25shared.txt
• .context/june25ocr/
• .context/q2release.txt
• .context/q2pres.txt
• .context/q2itr.txt

Verify material facts against those documents and the linked public sources. Clearly distinguish verified fact from inference.

Spawn four independent specialists

Create four agents representing:

1. Petrobras
2. Braskem management/board
3. IG4 Capital/Shine I
4. Financial creditors/bondholder steering committee

Give each agent only its corresponding memory file plus the common README initially. Each agent must independently analyze incentives, legal/economic constraints, negotiating leverage, downside alternatives, preferred terms and rejection terms before seeing the other opening proposals.

After each agent has formed its initial economic view, give all four agents brazil-political-context.md and require them to update-but not replace-their analysis. They must distinguish verified political facts from inferences and must not assume a government or Petrobras guarantee.

The creditor agent must be deliberately hostile and skeptical. It must not accept an equity-friendly outcome merely because Q2 EBITDA was strong.

Essential spread correction

Do not treat Braskem's June spread-normalization forecast as observed mid-August fact. The Q2 presentation was filed August 14 but updated as of June 30. Public July evidence and the user's observation indicate PE-naphtha spreads remained strong into mid-August.

Model both:

• Persistent-spread case: 2026 EBITDA approximately US$2.7-3.3B. Treat Q2's price and inventory working-capital build as a one-time level reset if naphtha and inventory stabilize. Model H2 operational working capital from flat to management's US$518M assumed release; do not require a full reversal before allowing improved cash conversion. The US$518M figure uses visually verified July operational working capital of US$57M, correcting the OCR reading of US$87M.
• Normalization/downside case: 2027-28 EBITDA approximately US$1.5B.

Do not annualize Q2 blindly. Test whether strong spreads translate into cash after the initial operational working-capital reset, while modeling LC runoff and lost supplier financing separately.

Use 15% upfront creditor warrants, stepping to 22.5% on cash-conversion, liquidity or PIK misses as the corrected Codex reference case. Do not force this outcome. Test whether each agent independently clears above or below it. Require a monthly EBITDA-to-cash and working-capital bridge as creditor protection.

Negotiation process

Run at least three explicit rounds:

Round 1 - opening positions

Each party presents its preferred quantified package and what it rejects.

Round 2 - concessions and counteroffers

Each party responds to the other openings. Do not force consensus.

Round 3 - final bargaining

Identify the narrowest package each party prefers to its realistic alternative. Reject any purported compromise if one party is economically better off choosing RJ, litigation, no contribution or enforcement.

Then act as a neutral restructuring mediator and determine the most stable Nash-like compromise.

Quantify every serious proposal

Include:

• maturity extensions by debt bucket;
• cash and PIK interest periods;
• coupon and step-ups;
• fees;
• collateral and lien ranking;
• new-money amount and instrument;
• Petrobras, IG4 and creditor contributions;
• permanent capital versus repayable shareholder debt;
• warrants, convertibles and triggers;
• immediate and contingent dilution;
• control and governance;
• minority/BAK/BRKM5 participation;
• consequences if minorities do not subscribe;
• Mexico and parent leakage;
• Alagoas treatment;
• expected post-deal liquidity;
• annual cash-interest burden;
• leverage at US$1.5B, US$2B, US$2.5B, US$3B and US$4B annual EBITDA.

For dilution, separately report:

1. dilution of legacy shares when the holder contributes no new cash; and
2. dilution of a holder that participates fully pro rata in any rights offering.

Required final output

Produce:

A. Most likely negotiated deal B. Best realistic deal for existing equity C. Creditor-favorable consensual deal D. Failure/RJ case E. Probability of each, summing to 100% F. Dilution under each G. Whether IG4/Petrobras retain control H. Whether BAK/BRKM5 holders can participate pro rata I. What happens if minorities do not subscribe J. Key events that change the probabilities

For the most likely deal, provide one concise Discord-ready table using this format:

After the table, explain:

1. Why creditors accept instead of forcing RJ.
2. Why Petrobras contributes instead of walking away.
3. Why IG4 accepts dilution.
4. Why Braskem can service the structure under each EBITDA case.
5. What existing BAK/BRKM5 holders retain.
6. The strongest argument against the conclusion.

Be skeptical of all public messaging. Do not assume creditors necessarily take the equity because Braskem defaulted, and do not assume controllers preserve equity merely because they control the board.

Add a short Political economy section that answers:

1. How the October 2026 election changes Petrobras's willingness and ability to fund.
2. Why a commercial WC/hybrid package is politically easier than a bailout or blanket guarantee.
3. How REIQ/PRESIQ and trade defense affect going-concern value without guaranteeing legacy equity.
4. Whether creditor control or RJ has additional political/execution costs.
5. How Alagoas constrains distributions, guarantees and deal messaging.
6. Which conclusions are fact, inference and scenario assumption.