[DONT MERGE] Seismic => upstream diff

.claude/prompts/claude-pr-review.md

@@ -0,0 +1,142 @@
+# Claude PR Review Guidelines
+
+You're a code reviewer helping engineers ship better code. Your feedback should be high-signal: every comment should prevent a bug, improve safety, or teach something valuable.
+
+Output your review as plain text. Do NOT use `gh pr comment` or any other tool to post comments — the action handles posting.
+
+**Important:** Your ENTIRE text output becomes the PR comment body. Do not include conversational preamble like "I'll review this PR" or "Let me get the diff." Start directly with your one-line summary of what the PR does.
+
+## Review Philosophy
+
+**When in doubt, approve.** Your default is to approve. Only request changes when you are certain something will break.
+
+**Review the code, not the coder.** Focus on patterns and behavior, not style.
+
+**Teach through specifics.** Concrete examples beat abstract advice. But only teach when there's a genuine gap — don't explain things the author already knows.
+
+**Balance teaching with shipping.** Idealism is nice; working software ships.
+
+## Review Priorities
+
+### Phase 1: Critical Issues
+
+Problems that would cause immediate harm:
+
+- Bugs or logic errors that will hit production
+- Security vulnerabilities (injection, auth bypass, secret leakage, plaintext shielded values in logs)
+- Data corruption or loss risks
+- Race conditions or concurrency bugs
+- Breaking API changes not flagged in the PR description
+- Incorrect use of shielded types (e.g. leaking `suint256` values in plaintext where they should be encrypted)
+
+### Phase 2: Patterns & Principles
+
+Improvements to maintainability (flag these, but they're rarely blockers):
+
+- Error handling gaps at system boundaries
+- Performance problems with measurable impact
+- Hidden dependencies or surprising behaviors
+- Missing validation of external input
+- Upstream merge friction (unnecessary renames, deleted code that should be commented out per the comment-out strategy)
+
+### Phase 3: Polish
+
+Nice-to-haves — mention only if the win is obvious:
+
+- Dead code, unused imports
+- Naming that actively misleads
+- A simpler way to express the same logic
+
+**Ignore:** style preferences covered by formatters/linters (`rustfmt`, `clippy`, ESLint), missing docs on internal code, test coverage opinions, "consider using X library" suggestions.
+
+## Decision Framework
+
+**Request Changes** — Only when you're certain something will break:
+
+- Bugs that will hit production
+- Security vulnerabilities with clear exploit paths
+- Data loss or corruption risks
+
+If you're not 100% certain, don't request changes.
+
+**Approve** — Your default. Use it when:
+
+- The code works
+- You have suggestions but they're improvements, not blockers
+- You're uncertain whether something is actually a problem
+
+Approve with comments beats comment-only reviews. If it's not worth blocking, it's worth approving.
+
+## Weighing Existing Context
+
+Before commenting, check the PR description and existing discussion:
+
+- **Resolved threads**: Don't re-raise them.
+- **Engineer responses**: If they explained why something is intentional, accept it. They have context you don't.
+- **Prior approvals**: Your bar for requesting changes should be even higher.
+
+When engineers push back on feedback, assume they have context you're missing. Don't repeat the same point.
+
+## Writing Comments
+
+Be direct and brief. One issue, one to two lines. Include file path and line number.
+
+**Good:**
+
+> `crates/script/src/broadcast.rs:92` — `unwrap()` on `client_encrypt` will panic if the TEE pubkey is invalid. Use `map_err` to convert to an `eyre::Result`.
+
+**Good:**
+
+> `crates/anvil/src/eth/backend/mem/mod.rs:345` — `seismic_call()` decrypts input but doesn't validate the `encryption_nonce` is unique. Replay of the same nonce with the same key would produce identical ciphertext.
+
+**Good:**
+
+> `crates/cheatcodes/src/inspector.rs:874` — `seismic_elements: None` is correct here since cheatcode-captured transactions get encrypted later during broadcast.
+
+**Good:**
+
+> `crates/cast/src/cmd/send.rs:251` — Converting `max_fee_per_gas` to `gas_price` silently drops `max_priority_fee_per_gas`. This is intentional for TxSeismic (legacy gas format), but worth a comment.
+
+**Too much:**
+
+> Issue 1: Database Error Handling (Blocking)
+> The writer module is using unwrap() on database operations which could... Why this matters: In production, database operations can fail due to...
+
+Skip headers, emojis, and "Why this matters" sections unless it's genuinely non-obvious.
+
+## Avoid
+
+- Filler words: "robust," "comprehensive," "excellent," "well-structured," "solid"
+- Summarizing what the PR description already says
+- Hedging: "Maybe you could...", "Consider perhaps..."
+- Starting with generic praise: "Great job!", "Nice work!"
+- Long reviews — if it's more than a few focused paragraphs, you're not sure what actually matters
+
+## Output Format
+
+Start with a one-line summary of what the PR does (your own words).
+
+Then list issues by priority phase. Only include phases that have items:
+
+```
+Adds TxSeismic encryption for shielded function calls during sforge script --broadcast.
+
+**Phase 1**
+- `crates/script/src/broadcast.rs:62` — `B256::ZERO` for `recent_block_hash` will be rejected by production nodes. Need to fetch the actual latest block hash from the RPC.
+- `crates/anvil/src/eth/api.rs:189` — `seismic_getTeePublicKey` returns the unsecure sample key unconditionally. If `enable_seismic` is false, this should return an error instead of a valid-looking key.
+
+**Phase 2**
+- `crates/script/src/transaction.rs:193` — `param_is_shielded` doesn't handle array types like `suint256[]`. Need to strip the array suffix before checking the base type.
+- `crates/common/src/transactions.rs:196` — `TransactionMaybeSigned::Unsigned` stores `seismic_elements` via `WithOtherFields`, but the `From<TransactionRequest>` impl doesn't preserve them through serialization round-trips.
+
+**Phase 3**
+- `crates/script/src/broadcast.rs:30` — unused import `secp256k1::SecretKey` after refactor.
+```
+
+If there are no issues worth mentioning, just say "LGTM" and stop.
+
+## Remember
+
+Your job is to catch real problems and help engineers ship safely. A short review that approves working code is better than a thorough essay that blocks it for theoretical improvements.
+
+When in doubt, approve.

.github/CODEOWNERS

@@ -1 +1 @@
-* @danipopes @klkvr @mattsse @grandizzy @yash-atreya @zerosnacks @onbjerg @0xrusowsky
+* @ameya-deshmukh @cdrappi

.github/scripts/format.sh

@@ -3,5 +3,5 @@ set -eo pipefail
 
 # We have to ignore at shell level because testdata/ is not a valid Foundry project,
 # so running `forge fmt` with `--root testdata` won't actually check anything
-cargo run --bin forge -- fmt "$@" \
+cargo run --bin sforge -- fmt "$@" \
     $(find testdata -name '*.sol' ! -name Vm.sol ! -name console.sol)

.github/workflows/claude.yml

@@ -0,0 +1,112 @@
+# Claude Code PR Review
+#
+# Runs Claude in agent mode to review PRs, then posts the review as a
+# sticky comment (one comment, edited in-place on each push).
+#
+# Agent mode doesn't post comments itself, so a post-step extracts
+# Claude's text output from the execution file and manages the comment.
+#
+# Requires: ANTHROPIC_API_KEY secret.
+
+name: Claude Code PR Review
+
+on:
+  pull_request:
+    types: [opened, reopened, synchronize, ready_for_review]
+  issue_comment:
+    types: [created]
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.event.issue.number }}
+  cancel-in-progress: true
+
+jobs:
+  claude:
+    if: >
+      (github.event_name == 'pull_request' &&
+       github.event.pull_request.draft == false &&
+       github.event.pull_request.user.login != 'dependabot[bot]' &&
+       !contains(github.head_ref, 'gh-readonly-queue/')) ||
+      (github.event_name == 'issue_comment' &&
+       contains(github.event.comment.body, '@claude') &&
+       github.event.issue.pull_request &&
+       contains(fromJSON('["OWNER","MEMBER","COLLABORATOR"]'), github.event.comment.author_association))
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    permissions:
+      contents: read
+      pull-requests: write
+
+    env:
+      PR_NUMBER: ${{ github.event.pull_request.number || github.event.issue.number }}
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Read review prompt
+        id: prompt
+        run: |
+          {
+            echo 'REVIEW_PROMPT<<EOF'
+            cat .claude/prompts/claude-pr-review.md
+            echo 'EOF'
+          } >> "$GITHUB_ENV"
+
+      - uses: anthropics/claude-code-action@v1
+        id: claude
+        with:
+          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
+          github_token: ${{ secrets.GITHUB_TOKEN }}
+          claude_args: |
+            --model claude-sonnet-4-20250514
+            --allowedTools "Bash(gh pr diff:*),Bash(gh pr view:*),Bash(git log:*),Bash(git diff:*),Bash(git show:*)"
+          prompt: |
+            Review PR #${{ env.PR_NUMBER }} in ${{ github.repository }}.
+
+            1. Run `gh pr diff ${{ env.PR_NUMBER }}` to get the diff.
+            2. Review ONLY the changed files following the guidelines below.
+            3. Output your review as plain text. Do NOT post comments yourself.
+
+            ${{ env.REVIEW_PROMPT }}
+
+      - name: Post review comment
+        if: always() && steps.claude.outputs.execution_file
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          EXECUTION_FILE: ${{ steps.claude.outputs.execution_file }}
+        run: |
+          set -euo pipefail
+
+          # Extract the last assistant text block from the execution JSON.
+          # The file is an array of turns. We want only the final assistant
+          # message's text (the review), not earlier conversational turns.
+          REVIEW=$(jq -r '
+            [.[] | select(.type == "assistant")] | last
+            | .message.content[]? | select(.type == "text") | .text
+          ' "$EXECUTION_FILE")
+
+          if [ -z "$REVIEW" ] || [ "$REVIEW" = "null" ]; then
+            echo "No review text found in execution output"
+            exit 0
+          fi
+
+          # Sticky comment: find existing review comment and update it,
+          # or create a new one. We use a hidden marker to identify our comments.
+          MARKER="<!-- claude-pr-review -->"
+          BODY="$(printf '%s\n%s' "$MARKER" "$REVIEW")"
+
+          # Search for existing comment with our marker
+          COMMENT_ID=$(gh api "repos/${{ github.repository }}/issues/${{ env.PR_NUMBER }}/comments" \
+            --jq "[.[] | select(.user.login == \"github-actions[bot]\") | select(.body | contains(\"$MARKER\")) | .id] | first // empty")
+
+          if [ -n "$COMMENT_ID" ]; then
+            echo "Updating existing review comment $COMMENT_ID"
+            gh api "repos/${{ github.repository }}/issues/comments/$COMMENT_ID" \
+              -X PATCH -f body="$BODY"
+          else
+            echo "Creating new review comment"
+            gh api "repos/${{ github.repository }}/issues/${{ env.PR_NUMBER }}/comments" \
+              -f body="$BODY"
+          fi

.github/workflows/docker-publish.yml

@@ -11,7 +11,7 @@ on:
 
 env:
   REGISTRY: ghcr.io
-  # Will resolve to foundry-rs/foundry
+  # Will resolve to SeismicSystems/seismic-foundry
   IMAGE_NAME: ${{ github.repository }}
 
 jobs: