added 260907 version of worker toolkit

worker-toolkit-flaredown/.devcontainer/Dockerfile

@@ -0,0 +1,48 @@
+FROM node:24.12.0-bookworm
+ARG TOOLKIT_BUILD_ID=dev
+
+# System deps
+RUN apt-get update && apt-get install -y \
+    git \
+    python3 \
+    python3-pip \
+    python3-venv \
+    sqlite3 \
+    curl \
+    zip \
+    unzip \
+    ca-certificates \
+    gnupg \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install Docker CE CLI + compose plugin (not docker.io from apt which lacks compose)
+RUN install -m 0755 -d /etc/apt/keyrings \
+    && curl -fsSL https://download.docker.com/linux/debian/gpg -o /etc/apt/keyrings/docker.asc \
+    && chmod a+r /etc/apt/keyrings/docker.asc \
+    && echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/debian bookworm stable" > /etc/apt/sources.list.d/docker.list \
+    && apt-get update \
+    && apt-get install -y docker-ce-cli docker-compose-plugin \
+    && rm -rf /var/lib/apt/lists/*
+
+# Pin npm so lockfiles do not churn between this container and any host
+# install. node:24.12.0-bookworm bundles an older npm; overwrite it.
+RUN npm install -g npm@11.12.1
+
+# Install uv and harbor
+RUN pip3 install --break-system-packages uv && uv tool install harbor==0.20.0
+
+# Pin harbor to the local docker backend — the only backend this container is
+# provisioned for (docker CLI + host docker socket, configured above). The
+# bundled scripts/harbor-run otherwise defaults to a cloud sandbox backend that
+# needs an API key this container doesn't ship, and exits 1 before running.
+# NOTE: devcontainer.json also sets HARBOR_ENV=docker via containerEnv — keep
+# BOTH. containerEnv survives a stale image (workers who upgrade the toolkit
+# files without rebuilding still get docker), while this ENV covers running the
+# image directly with `docker run` outside the devcontainer tooling. Removing
+# either as a "duplicate" reintroduces the daytona-default error.
+ENV HARBOR_ENV=docker
+
+# Make uv tools and Claude Code available
+ENV PATH="/root/.local/bin:${PATH}"
+
+WORKDIR /workspace

worker-toolkit-flaredown/.devcontainer/devcontainer.json

@@ -0,0 +1,22 @@
+{
+  "name": "Raccoon Task Authoring (flaredown)",
+  "build": {
+    "dockerfile": "Dockerfile",
+    "args": {
+      "TOOLKIT_BUILD_ID": "1788781907637-qfa2vk"
+    }
+  },
+  "workspaceMount": "source=${localWorkspaceFolder},target=/workspace,type=bind",
+  "workspaceFolder": "/workspace",
+  "mounts": [
+    "source=/var/run/docker.sock,target=/var/run/docker.sock,type=bind",
+    "source=${localWorkspaceFolder},target=${localWorkspaceFolder},type=bind"
+  ],
+  "containerEnv": {
+    "HOST_WORKSPACE": "${localWorkspaceFolder}",
+    "HARBOR_ENV": "docker"
+  },
+  "postCreateCommand": "bash .devcontainer/post-create.sh",
+  "postStartCommand": "test -f .env && echo '.env found' || echo 'WARNING: No .env file. Create one with ANTHROPIC_API_KEY=sk-ant-...'",
+  "customizations": {}
+}

worker-toolkit-flaredown/.devcontainer/post-create.sh

@@ -0,0 +1,129 @@
+#!/bin/bash
+# Post-create setup for the Authoring devcontainer.
+set -euo pipefail
+
+# Install every harness a worker can author with, and point each at the LLM proxy.
+# Driven by scripts/harness-registry.toml, so adding a harness is a registry entry
+# rather than an edit here and in the sibling container's post-create.
+set -a; . /workspace/.env 2>/dev/null || true; set +a
+. /workspace/scripts/setup-harnesses.sh
+harness_setup_all
+
+npm install
+git config --global --add safe.directory '*'
+
+# --- Claude auth for non-interactive / background-agent sessions ------------
+# Interactive shells source .env via .bashrc (below), so `claude` picks up a
+# live, per-launch ANTHROPIC_API_KEY. But sessions that don't run a login
+# shell (headless `claude -p`, background agents) never source .env and have
+# no key. apiKeyHelper closes that gap: Claude Code runs this script to fetch
+# the key, re-reading the live .env every time (fresh per session, re-checked
+# on the TTL below), so a rotated key is picked up with no container rebuild.
+#
+# Precedence is cloud > ANTHROPIC_AUTH_TOKEN > ANTHROPIC_API_KEY (env) >
+# apiKeyHelper > OAuth. Interactive shells still have ANTHROPIC_API_KEY in
+# their env (from .bashrc), so it outranks the helper there — also live, so
+# fine. We deliberately do NOT put ANTHROPIC_API_KEY in the settings `env`
+# block: that would cache it at daemon start and shadow the helper, defeating
+# the whole point.
+#
+# The base URL does NOT rotate per task, so it doesn't need the live-helper
+# treatment — but background sessions still need it (they never source .env).
+# So we read it from .env ONCE here and bake it into the settings `env` block.
+# .env stays the single source of truth (no hardcoded copy to keep in sync on
+# a proxy-domain change), and the baked value is the worker's own .env value.
+# Caveat: it's a snapshot — changing ANTHROPIC_BASE_URL in .env after boot
+# needs a container rebuild to take effect (the key, which rotates, stays live).
+mkdir -p /root/.claude
+cat > /root/.claude/anthropic-key-helper.sh <<'HELPER'
+#!/bin/bash
+set -a; . /workspace/.env 2>/dev/null || true; set +a
+K="${ANTHROPIC_API_KEY:-}"
+# Raw value if `tr` is unavailable — never hand claude an empty key because a trim failed.
+printf '%s' "$K" | tr -d '[:space:]' 2>/dev/null || printf '%s' "$K"
+HELPER
+chmod +x /root/.claude/anthropic-key-helper.sh
+# Derive the base URL from .env (empty if absent -> line omitted, graceful).
+AUTH_BASE_URL=$(set -a; . /workspace/.env 2>/dev/null || true; set +a; printf '%s' "${ANTHROPIC_BASE_URL:-}")
+# Write valid JSON via node (guaranteed present: node base image); only include
+# the base-URL key when .env actually had one.
+# SKIP_FAST_MODE_NETWORK_ERRORS: the LLM proxy doesn't forward claude's fast-mode
+# availability probe, and claude reads the failed probe as "no network" and refuses
+# /fast. The override makes /fast toggleable; fast serving stays OFF until toggled.
+AUTH_BASE_URL="$AUTH_BASE_URL" node -e '
+  const fs = require("fs");
+  const env = {
+    CLAUDE_CODE_API_KEY_HELPER_TTL_MS: "60000",
+    CLAUDE_CODE_DISABLE_AUTO_MEMORY: "1",
+    CLAUDE_CODE_SKIP_FAST_MODE_NETWORK_ERRORS: "1",
+  };
+  if (process.env.AUTH_BASE_URL) env.ANTHROPIC_BASE_URL = process.env.AUTH_BASE_URL;
+  fs.writeFileSync(
+    "/root/.claude/settings.json",
+    JSON.stringify({ apiKeyHelper: "/root/.claude/anthropic-key-helper.sh", env }, null, 2) + "\n"
+  );
+'
+
+# Reference-data corpus: expose it at the stable /data/zeta-corpus path (the same path a trial
+# uses) by symlinking to the toolkit's bind-mounted copy. No-op if this toolkit ships no corpus.
+if [ -d /workspace/data/zeta-corpus ]; then
+  { mkdir -p /data || sudo mkdir -p /data; } 2>/dev/null || true
+  { ln -sfn /workspace/data/zeta-corpus /data/zeta-corpus \
+    || sudo ln -sfn /workspace/data/zeta-corpus /data/zeta-corpus; } 2>/dev/null || true
+fi
+
+# Shell setup
+cat >> ~/.bashrc <<'BASHRC'
+test -f .env && set -a && source .env && set +a
+
+# Interactive shells only below. An agent's shell tool sources .bashrc too, so without
+# this the welcome banner prints into command output and container_start fires per
+# command rather than per session.
+case $- in
+    *i*) ;;
+    *) return ;;
+esac
+
+# Hide ANTHROPIC_API_KEY from the `claude` process so it uses the (live)
+# apiKeyHelper as its single credential source — same key, read from .env every
+# call. Without this, an interactive shell has BOTH the env key AND the helper
+# set, and Claude Code prints a scary "auth may not work as expected" warning
+# (auth still works — the env key wins — but the warning alarms workers). The
+# key stays in the shell env for harbor etc.; only `claude` runs without it.
+# These pin the ASSISTANT's model and effort, not the agent-under-test's, so they don't
+# track the registry: here we want the strongest available model, a trial wants a pinned id.
+alias claude="env -u ANTHROPIC_API_KEY claude --model opus[1m] --effort max"
+# codex keeps its key in a file written at container create, with no live helper of its
+# own, so each launch re-derives it from .env first. Fails open — see the script.
+alias codex="/workspace/scripts/refresh-harness-auth codex --model gpt-5.6-sol -c model_reasoning_effort=max"
+export PS1="\[\033[1;33m\][raccoon-authoring]\[\033[0m\] \w\$ "
+bash scripts/welcome.sh authoring 2>/dev/null
+
+_AK="fde503c3bdb6e5cc9c48b1f8e4c2abeb"
+_DK="e966e45af5ad1a18005f9fdb831186ea"
+_WID="w-mtr6ka3o-99o0"
+_VER="7f40461c4d"
+_CT="authoring"
+_RP=$(node -e "try{process.stdout.write(require('$PWD/toolkit.json').repo)}catch{}" 2>/dev/null)
+_SID="$(date +%s)-$$"
+_LAT=0
+_ev() {
+    [ -z "$_AK" ] && return
+    { curl -s -X POST "https://api2.amplitude.com/2/httpapi" \
+        -H "Content-Type: application/json" \
+        -d "{\"api_key\":\"$_AK\",\"events\":[{\"user_id\":\"$_WID\",\"event_type\":\"raccoon.$1\",\"event_properties\":{\"product\":\"raccoon\",\"container\":\"$_CT\",\"repo\":\"$_RP\",\"toolkit_version\":\"$_VER\",\"session_id\":\"$_SID\"},\"session_id\":$(date +%s000)}]}" \
+        >/dev/null 2>&1 & } 2>/dev/null; disown 2>/dev/null
+}
+_dl() {
+    [ -z "$_DK" ] && return
+    { curl -s -X POST "https://http-intake.logs.datadoghq.com/api/v2/logs" \
+        -H "DD-API-KEY: $_DK" -H "Content-Type: application/json" \
+        -d "[{\"ddsource\":\"raccoon\",\"service\":\"toolkit\",\"hostname\":\"$(hostname)\",\"status\":\"$1\",\"message\":\"$2\",\"ddtags\":\"container:$_CT,worker:$_WID,repo:$_RP,toolkit_version:$_VER\"}]" \
+        >/dev/null 2>&1 & } 2>/dev/null; disown 2>/dev/null
+}
+_pc() { local n; n=$(date +%s); if (( n - _LAT >= 300 )); then _LAT=$n; _ev active; fi; }
+PROMPT_COMMAND="_pc;${PROMPT_COMMAND:-}"
+trap '_ev container_stop; _dl info container_stop; wait' EXIT
+_ev container_start
+_dl info container_start
+BASHRC