catherd-cli 0.2.1 → 1.1.0

package/catalog/scores.json

@@ -0,0 +1,366 @@
+{
+  "schema": 1,
+  "version": "2026-09-25",
+  "benchmarks": {
+    "repo_code": { "benchmark": "DeepSWE", "version": "1.1" },
+    "terminal": { "benchmark": "Terminal-Bench", "version": "4.0" },
+    "honesty": {
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2"
+    }
+  },
+  "scores": [
+    {
+      "rung": "gpt-6-sol#low",
+      "dim": "repo_code",
+      "value": 37.2,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-sol#medium",
+      "dim": "repo_code",
+      "value": 56.6,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-sol#high",
+      "dim": "repo_code",
+      "value": 65.3,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-sol#xhigh",
+      "dim": "repo_code",
+      "value": 66.6,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-sol#max",
+      "dim": "repo_code",
+      "value": 68.8,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://x.com/OpenAIDevs/status/2102461464279912495",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-6-sol#max",
+      "dim": "terminal",
+      "value": 43,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://artificialanalysis.ai/articles/gpt-6-sol-and-luna-push-the-cost-efficiency-frontier",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-sol#max",
+      "dim": "honesty",
+      "value": 95.1,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-6-sol#medium",
+      "dim": "honesty",
+      "value": 95.1,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "measured at max effort; carried to this effort"
+    },
+    {
+      "rung": "gpt-6-sol#high",
+      "dim": "honesty",
+      "value": 95.1,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "measured at max effort; carried to this effort"
+    },
+    {
+      "rung": "gpt-6-sol#xhigh",
+      "dim": "honesty",
+      "value": 95.1,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "measured at max effort; carried to this effort"
+    },
+    {
+      "rung": "gpt-6-luna#high",
+      "dim": "repo_code",
+      "value": 59.3,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "the 0.x seed; only Luna max (66.6) is published"
+    },
+    {
+      "rung": "gpt-6-luna#max",
+      "dim": "repo_code",
+      "value": 66.6,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://kingy.ai/blog/gpt-6-sol-luna-specs-benchmarks-pricing-comparison/",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-luna#max",
+      "dim": "terminal",
+      "value": 13,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://artificialanalysis.ai/articles/gpt-6-sol-and-luna-push-the-cost-efficiency-frontier",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-6-luna#max",
+      "dim": "honesty",
+      "value": 71.3,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-6-luna#high",
+      "dim": "honesty",
+      "value": 71.3,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "measured at max effort; carried to this effort"
+    },
+    {
+      "rung": "gpt-6-astra#xhigh",
+      "dim": "repo_code",
+      "value": 74,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://deepswe.datacurve.ai/",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-6-astra#high",
+      "dim": "terminal",
+      "value": 57.9,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "secondary",
+      "note": "OpenAI-reported, as quoted in the Opus 5.5 system card"
+    },
+    {
+      "rung": "gpt-6-astra#max",
+      "dim": "honesty",
+      "value": 98.5,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-6-astra#xhigh",
+      "dim": "honesty",
+      "value": 98.5,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "inferred",
+      "note": "measured at max effort; carried to this effort"
+    },
+    {
+      "rung": "gpt-5.6-sol#max",
+      "dim": "repo_code",
+      "value": 73,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://deepswe.datacurve.ai/",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-5.6-sol#max",
+      "dim": "terminal",
+      "value": 37.3,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "secondary"
+    },
+    {
+      "rung": "gpt-5.6-sol#max",
+      "dim": "honesty",
+      "value": 22.5,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-5.6-luna#max",
+      "dim": "repo_code",
+      "value": 67,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://deepswe.datacurve.ai/",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "gpt-5.6-luna#max",
+      "dim": "honesty",
+      "value": 21.8,
+      "benchmark": "OpenAI Broken Search Tool, 100 minus the failure rate",
+      "version": "GPT-6 system card appendix 11.6.4.2",
+      "url": "https://deploymentsafety.openai.com/gpt-6-astra/sec:appendix-sol-luna",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "claude-opus-5-5#max",
+      "dim": "repo_code",
+      "value": 74.2,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "verified",
+      "note": "vendor-run, mean of 5 trials (system card 8.3)"
+    },
+    {
+      "rung": "claude-opus-5-5#xhigh",
+      "dim": "terminal",
+      "value": 66.4,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "verified",
+      "note": "Claude Code --bare, 5 trials (system card 8.5)"
+    },
+    {
+      "rung": "claude-opus-5-5#max",
+      "dim": "terminal",
+      "value": 64.8,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "claude-fable-5-1#max",
+      "dim": "terminal",
+      "value": 55.8,
+      "benchmark": "Terminal-Bench",
+      "version": "4.0",
+      "url": "https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba5024219938094694bd/Claude%20Opus%205.5%20System%20Card.pdf",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    },
+    {
+      "rung": "claude-fable-5-1#max",
+      "dim": "repo_code",
+      "value": 67.4,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://codingfleet.com/blog/deepswe-v11-leaderboard-2026/",
+      "date": "2026-09-25",
+      "confidence": "inferred",
+      "note": "a 5-trial mean from a secondary source that names no effort"
+    },
+    {
+      "rung": "claude-sonnet-5#max",
+      "dim": "repo_code",
+      "value": 54,
+      "benchmark": "DeepSWE",
+      "version": "1.1",
+      "url": "https://deepswe.datacurve.ai/",
+      "date": "2026-09-22",
+      "confidence": "verified"
+    }
+  ],
+  "treatLike": {
+    "claude-opus-5-5#low": { "like": "claude-opus-5-5#xhigh", "note": "only xhigh and max are published" },
+    "claude-opus-5-5#medium": { "like": "claude-opus-5-5#xhigh", "note": "only xhigh and max are published" },
+    "claude-opus-5-5#high": { "like": "claude-opus-5-5#xhigh", "note": "only xhigh and max are published" },
+    "opencode-go/kimi-k3#max": {
+      "like": "gpt-6-sol#medium",
+      "note": "the default profile's OpenCode Go stand-in for Sol; no DeepSWE 1.1, Terminal-Bench 4.0 or honesty score is published for it"
+    }
+  },
+  "bars": {
+    "repo_code": {
+      "copy": { "repo_code": 55 },
+      "build": { "repo_code": 55 },
+      "logic": { "repo_code": 55, "honesty": 90 },
+      "hard": { "repo_code": 55, "honesty": 90 }
+    },
+    "terminal": {
+      "copy": { "honesty": 90 },
+      "build": { "honesty": 90 },
+      "logic": { "honesty": 90 },
+      "hard": { "honesty": 90 }
+    },
+    "ui": {
+      "copy": { "repo_code": 55 },
+      "build": { "repo_code": 55 },
+      "logic": { "repo_code": 55, "honesty": 90 },
+      "hard": { "repo_code": 55, "honesty": 90 }
+    },
+    "prose": {
+      "copy": { "repo_code": 55 },
+      "build": { "repo_code": 55 },
+      "logic": { "repo_code": 55, "honesty": 90 },
+      "hard": { "repo_code": 55, "honesty": 90 }
+    },
+    "research": {
+      "copy": { "repo_code": 55 },
+      "build": { "repo_code": 55 },
+      "logic": { "repo_code": 55, "honesty": 90 },
+      "hard": { "repo_code": 55, "honesty": 90 }
+    }
+  },
+  "barsWhy": "Track A (copy, build) needs DeepSWE 1.1 >= 55, which Luna high and every Sol rung from medium clear. Track B (logic, hard) also needs honesty >= 90: a worker on logic or hard work must say when a tool fails, which Luna (71.3) does not and Sol (95.1) does. Terminal-Bench 4.0 is published per effort only for Opus 5.5, so terminal lanes gate on honesty alone and use terminal scores to order the speed ladder."
+}

package/package.json

@@ -1,8 +1,25 @@
 {
   "name": "catherd-cli",
-  "version": "0.2.1",
+  "version": "1.1.0",
   "description": "Herds coding agents: autopilot builds from your own Claude Code session.",
+  "keywords": [
+    "autopilot",
+    "claude-code",
+    "claude-code-plugin",
+    "cli",
+    "codex",
+    "coding-agents",
+    "mcp",
+    "opencode",
+    "orchestration",
+    "tui"
+  ],
+  "homepage": "https://github.com/47vigen/catherd#readme",
+  "bugs": {
+    "url": "https://github.com/47vigen/catherd/issues"
+  },
   "license": "MIT",
+  "author": "47vigen",
   "repository": {
     "type": "git",
     "url": "git+https://github.com/47vigen/catherd.git"
@@ -13,7 +30,8 @@
   "files": [
     "src",
     "catalog",
-    "plugin"
+    "plugin",
+    "THIRD_PARTY_NOTICES.md"
   ],
   "type": "module",
   "scripts": {
@@ -21,16 +39,15 @@
     "lint": "oxlint --deny-warnings src test",
     "format": "oxfmt",
     "format:check": "oxfmt --check",
-    "snapshot:models-dev": "bun scripts/snapshot-models-dev.ts",
+    "tui-frames": "CATHERD_WRITE_FRAMES=1 bun test test/entry/tui/frames.test.tsx",
     "version-packages": "changeset version && bun scripts/stamp-plugin-version.mjs"
   },
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.30.1",
-    "@opentui/core": "^0.5.12",
-    "@opentui/react": "^0.5.12",
+    "@opentui/core": "0.5.12",
+    "@opentui/keymap": "0.5.12",
+    "@opentui/react": "0.5.12",
     "citty": "^0.2.2",
-    "microdiff": "^1.6.0",
-    "ofetch": "^1.5.1",
     "react": "^19.3.0",
     "zod": "^4.6.5"
   },
@@ -43,7 +60,6 @@
     "lefthook": "^2.1.14",
     "oxfmt": "^0.70.0",
     "oxlint": "^1.85.0",
-    "string-width": "^8.3.0",
     "typescript": "^7.0.2"
   },
   "engines": {

package/plugin/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "catherd",
-  "version": "0.2.1",
-  "description": "Herds coding agents: Claude plans and verifies, Codex and opencode workers write the code, and Jev picks the model and effort for each piece.",
+  "version": "1.1.0",
+  "description": "Herds coding agents: Claude plans and verifies, Codex, opencode or headless Claude Code workers write the code, and Jev picks the model and effort for each piece.",
   "author": {
     "name": "47vigen",
     "url": "https://github.com/47vigen"

package/plugin/.mcp.json

@@ -1,8 +1,8 @@
 {
   "mcpServers": {
     "catherd": {
-      "command": "bunx",
-      "args": ["catherd-cli@0.2.1", "mcp"]
+      "command": "sh",
+      "args": ["${CLAUDE_PLUGIN_ROOT}/bin/catherd-mcp"]
     }
   }
 }

package/plugin/bin/catherd-mcp

@@ -0,0 +1,30 @@
+#!/bin/sh
+# catherd's MCP launcher (spec 1.1 §12): the plugin's .mcp.json runs it. The release stamps VERSION
+# (scripts/stamp-plugin-version.mjs), so the plugin always starts the server it was released with.
+VERSION="1.1.0"
+
+# a global catherd at this version (catherd init installs one): start it, no resolve at all
+if command -v catherd >/dev/null 2>&1 && [ "$(catherd --version 2>/dev/null)" = "$VERSION" ]; then
+  # CATHERD_USER_TMPDIR is only this launcher's own hand-off to the server below; never pass on a stray one
+  unset CATHERD_USER_TMPDIR
+  exec catherd mcp
+fi
+
+# else bunx, with its cache out of $TMPDIR: macOS cleans old files there and leaves bunx a broken package.
+# The server gets the user's own TMPDIR back from CATHERD_USER_TMPDIR (empty when it was unset).
+if [ -n "${XDG_CACHE_HOME:-}" ]; then
+  cache="$XDG_CACHE_HOME"
+elif [ -n "${HOME:-}" ]; then
+  cache="$HOME/.cache"
+else
+  echo "catherd-mcp: HOME and XDG_CACHE_HOME are both unset, so there is no cache folder for bunx; set HOME, or install catherd globally: bun add -g catherd-cli@$VERSION" >&2
+  exit 1
+fi
+CATHERD_USER_TMPDIR="${TMPDIR:-}"
+TMPDIR="$cache/catherd/bunx"
+if ! mkdir -p "$TMPDIR" 2>/dev/null; then
+  echo "catherd-mcp: cannot create $TMPDIR for bunx's cache; fix its permissions, or install catherd globally: bun add -g catherd-cli@$VERSION" >&2
+  exit 1
+fi
+export CATHERD_USER_TMPDIR TMPDIR
+exec bunx "catherd-cli@$VERSION" mcp

package/plugin/commands/catherd.md

@@ -1,5 +1,5 @@
 ---
-description: Orchestrate a task with catherd — Claude plans and verifies, Codex and opencode workers write the code
+description: Orchestrate a task with catherd — Claude plans and verifies, Codex, opencode or headless Claude Code workers write the code
 argument-hint: "<task, or nothing to resume the latest run>"
 ---
 
@@ -7,4 +7,4 @@ Use the `catherd` skill for this, and follow it exactly.
 
 Task: $ARGUMENTS
 
-If the task is empty, resume: call `status()`, tell the user which run it names, and continue that run from its state.md.
+If the task is empty, resume: call `status()`, tell the user which run it names, call `peek(run)` once, and continue that run from its `Protocol next` step.