agentforce314 · ericleepi314 · Jul 29, 2026 · Jul 29, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -7,6 +7,52 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [1.3.0] - 2026-07-29
+
+### Added
+
+- Added the `claude-opus-5` model, and wired `--effort` end to end (the
+  interactive `/effort` path on Anthropic was also fixed) (#722, #746).
+- Added bounded persistent memory with a background self-improvement review
+  fork (#731).
+- Added a VS Code extension (`vscode-extension/clawcodex-vscode`) driving the
+  agent-server over stdio, with interrupt/suggestion/chosen-updates contracts
+  (#727).
+- Added image-paste input: paste an image at the prompt and it attaches as a
+  content block, shown as an `[Image #N]` chip that un-attaches on demand
+  (#761, #762).
+- Added a Harbor eval harness (`eval/harbor/`) with clawcodex, openclaude, and
+  latest-Claude-Code subscription adapters for terminal-bench 2.0/2.1 three-way
+  comparisons, ATIF trajectory emission, and per-step token/cost accounting
+  (#720, #724, #725, #736–#738).
+
+### Changed
+
+- Renamed the project context file `CLAUDE.md` → `CLAWCODEX.md` (clean break,
+  no fallback) (#732).
+- Tuned the agent prompt and headless harness for reliability and parity with
+  the reference: restored dropped task-tool skip conditions, parallel-tool
+  guidance, and dropped instruction qualifiers; deferred nonessential initial
+  tools; and recovered trials lost to empty turns and transport drops
+  (#743–#745, #747–#754).
+- Reworked the TUI header box to the reference's element allocation (#764).
+- Honor runtime 1M-context model limits end to end (#730).
+- Activated `--allowedTools`/`--disallowedTools`, which were a silent no-op
+  (#739).
+
+### Fixed
+
+- Ping-aware stream watchdog to stop spurious `NonZeroAgentExitCodeError` on
+  large-context agentic runs (#734).
+- Recover trials lost to empty turns, transport drops, and a headless-only
+  tool; preserve thinking and reduce headless overhead (#735, #742).
+- Enable interactive TaskV2 and harden Bash execution (#741).
+- Retry transient transport failures instead of aborting the run, and stop
+  `Read` from leaking image base64 into context (#757, #760).
+- Cap `mcp` below 2.0 (2.0.0 removed `mcp.client.websocket`) (#763).
+- Correct the DeepSeek V4 `max_output_tokens` to the documented 384K (#758).
+- Nudge off a stalled background task instead of polling forever (#759).
+
 ## [1.2.1] - 2026-07-16
 
 ### Changed

diff --git a/README.md b/README.md
@@ -119,6 +119,7 @@ The `session`, `settings`, and `env` blocks are optional — sensible defaults a
 
 ## 📰 News
 
+- **2026-07-29 (v1.3.0):** **ClawCodex scores 80.9% on Terminal-Bench 2.1 — a top-tier open-source result on Opus 5 (#720–#725, #747–#754)** — running headless on `claude-opus-5` at `effort=xhigh`, ClawCodex solved **72 of 89** Terminal-Bench 2.1 tasks: **80.9% pass@1** on a single run. On the [public 2.1 leaderboard](https://www.tbench.ai/leaderboard/terminal-bench/2.1) (k=5 averages) that would slot **around third** — behind Claude Code / Fable 5 (83.8%) and Codex / GPT-5.5 (83.1%), statistically level with the 79–80% cluster, and **ahead of Claude Code on Opus 4.8 (78.9%) and Sonnet 5 (74.6%)**. Getting there was open, unglamorous parity work: a Harbor eval adapter (`eval/harbor/`) for three-way ClawCodex-vs-openclaude-vs-Claude-Code runs (#720, #724, #725), then a run of prompt- and reliability-parity fixes — restored task-tool skip conditions and parallel-tool guidance, deferred nonessential initial tools, and recovery of trials lost to empty turns and transport drops (#747–#754). **Also in v1.3.0:** `claude-opus-5` support with an interactive `/effort` fix (#746), bounded persistent memory with a background self-improvement review (#731), a VS Code extension driving the agent-server over stdio (#727), image-paste input with an `[Image #N]` un-attach chip (#761, #762), the `CLAUDE.md → CLAWCODEX.md` context-file rebrand (#732), and transport-retry hardening (#757, #760). Stated plainly: this is a single k=1 pass (binomial 1σ ±4.2pp) against the board's k=5 ± ~1.2pp averages, benchmarked on `main` at #756 (before the v1.3.0 tag), so read it as directional rather than a ranked submission.
 - **2026-07-13:** **`/eco` token compression — -80% Bash-output tokens, measured, now a headline (#708, #712)** — a new session toggle compresses the model-bound rendering of every Bash result with deterministic filters ported from [RTK](https://github.com/rtk-ai/rtk)'s method set: failure-focused test summaries (kept error lines are never rewritten), `git`/`pip`/`npm` ceremony stripping, log dedup with `[×N]` counts, and a recoverable head-cap — all behind a **never-worse** guard, with every lossy compression teeing the full output to disk behind a runnable recovery hint (#708). A reproducible benchmark (`eval/eco/`) replays 27 real command outputs through the exact production pipeline and counts real tokenizer tokens: **92,989 → 17,767 (-80%)** corpus-wide, -88% on filter hits, plus an honestly conservative recompute of RTK's own 30-minute-session model (-19% under their averaged assumptions — real sessions are fat-tailed) (#712). Full tables: the [`/eco` section](#eco-benchmark) and [`eval/eco/results/`](eval/eco/results/results.md).
 - **2026-07-12 (v1.1.0):** **ClawCodex v1.1.0 — run OpenAI *and* Claude models on your subscription, not metered API billing** — the headline of 1.1.0 is **subscription auth for the two biggest model families**, so you can point ClawCodex at a plan you already pay for. **Sign in with ChatGPT (#698):** `clawcodex login → openai → subscription` (browser, device-code, or import from an existing Codex CLI login) routes requests through the ChatGPT Codex backend's Responses API — `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, and `gpt-5.3-codex-spark` on your Plus/Pro allowance, with encrypted-reasoning replay across turns and **$0** metered cost. **Claude Pro/Max (#697):** `clawcodex login → anthropic → subscription` connects a Claude subscription via OAuth (PKCE) with automatic token refresh and the same $0 accounting; follow-ups repaired the login after Anthropic moved its OAuth endpoints to `platform.claude.com` (#702) and stopped sending adaptive thinking to models that don't support it (#699). A configured API key always wins, and subscription usage reports `billing_mode: subscription`. **More models:** a Meta (`api.meta.ai`) provider with the 1M-context `muse-spark-1.1` reasoning model (#692) and refreshed MiniMax parameters (#696). **Workflow & TUI:** `/plan` mode with implicit plan-mode entry/exit (#676), `--worktree/-w` session isolation for parallel runs in separate git worktrees (#672), the `/memory` picker + `$EDITOR` spawn (#693), config/state directories rebranded `.claude → .clawcodex` with a one-time migration (#678), `/logo` startup color schemes (#677), plus TUI polish — Tab accepts the suggested placeholder (#690), past inputs get the Claude-Code highlight band (#691), clickable agent URLs (#694), and a per-terminal link-open affordance (#701). **Quality:** semantic tool-input coercion with parity validation errors (#700) and looser, Claude-Code-faithful permission granting (#673).
 - **2026-07-07:** **`/loop` scheduled tasks now actually fire — full port of Claude Code's session-scoped scheduler (#680)** — the bundled `/loop` skill finally has a real engine behind it: a new `src/scheduled_tasks` module parses standard 5-field cron expressions and fires due prompts **between turns** from the agent-server's idle poll. `CronCreate`/`CronList`/`CronDelete` register real firing jobs (8-char IDs, 50-job cap, deterministic jitter, 7-day recurring expiry with one final fire), and the new **`ScheduleWakeup`** tool drives self-paced `/loop` mode — the model picks each next delay (1 min–1 hr), `stop: true` ends the loop, and a ~20-minute fallback wakeup catches iterations that forget to reschedule. Typed skill slash commands now reach the backend (new `skill_command` control), so `/loop 5m check ci` works from the composer with completion + argument hint; the TUI shows a live countdown indicator (`⟳ loop wakeup in 2m 14s · ⏰ 1 scheduled`) and **Esc while idle stops a waiting loop**. `/clear` drops session tasks, `--resume` restores unexpired ones, `CLAWCODEX_DISABLE_CRON=1` disables the scheduler. 117 new tests; verified live over stdio NDJSON and a real PTY TUI drive (typed dispatch → CronCreate → a real wakeup fire between turns → Esc-stop).
@@ -128,7 +129,6 @@ The `session`, `settings`, and `env` blocks are optional — sensible defaults a
 - **2026-06-24 (v0.6.0):** **ClawCodex v0.6.0 — interactive TUI REPL parity** — a batch of input ports brings the Python REPL to parity with the ink reference: a working slash-command menu (execute / complete / filter like the ink REPL), the sparkle spinner with a live token + elapsed busy row, context-aware prompt footer hints (interrupt / bash / grammar), the `?` shortcuts help panel, an `@` file-mention dropdown with in-place splice, double-press Ctrl+C / Ctrl+D to exit, Ctrl+R history search + double-Esc clear-draft, a `[Pasted text #N +K lines]` large-paste placeholder, and the completed command queue (drain queued prompts + dim preview). Login docs now list all 25 providers (#383).
 - **2026-06-23:** **One-click installer** — `curl -fsSL https://clawcodex.app/install.sh | bash` installs uv (no sudo), provisions Python 3.10+, clones to `~/.clawcodex`, creates a lock-pinned venv, and registers `clawcodex` on PATH; ships status / doctor / verify / update / uninstall subcommands, is safe to re-run, and works on macOS / Linux / WSL.
 - **2026-06-21:** **18 new LLM providers — the registry grows 7 → 25 (#377)** — a data-driven `ProviderSpec` registry adds 18 OpenAI-compatible backends (nvidia-nim, fireworks, together, moonshot/Kimi, novita, siliconflow, deepinfra, stepfun, arcee, huggingface, volcengine, xiaomi-mimo, atlascloud, wanjie-ark, plus local ollama / vllm / sglang) alongside the hand-written providers; alias-aware config resolution, standard env-var key fallback (e.g. `TOGETHER_API_KEY`), and keyless local servers.
-- **2026-06-18:** **DeepSeek prefix-cache exploitation — a HUGE token-cost win (#363)** — ClawCodex now keeps its request prefix **byte-stable** across turns so DeepSeek's automatic prompt-prefix cache covers the entire `system + tools + history` span. Per-request-volatile sections (env, the mutable `MEMORY.md` body, plan-mode, etc.) are relocated to a trailing `<system-reminder>` *after* the conversation history, so the cached prefix never breaks even when memory/env change. We also register DeepSeek's **1M-token context window**, map its prompt-cache usage onto the Anthropic `cache_read_input_tokens` convention, and surface a per-model **prompt-cache hit-rate** + cost in `/cost`. **Why this is enormous — the token economics:** Claude Fable 5 runs **$10 / $50** per 1M input/output tokens, while **DeepSeek-V4-Pro is just $0.435 / $0.87** — already **~23× cheaper on input** and **~57× cheaper on output**. And because **cache-hit input is billed at only 10%** of the normal input rate, the long, context-heavy sessions that agentic coding actually produces pay just **~$0.0435 per 1M input tokens** — roughly **230× cheaper than Fable 5 input**. The token efficiency ClawCodex unlocks here is **HUGE**. Everything is gated to the `deepseek` provider — every other provider's request is byte-for-byte unchanged. Follow-up: truncated tool-call argument JSON is now best-effort recovered in the shared OpenAI-compatible layer, so an interrupted DeepSeek stream keeps its partial tool args instead of dropping them to `{}` (#364).
 📚 Older items have moved to the full **[News archive](docs/NEWS.md)**.
 
 ***

diff --git a/docs/NEWS.md b/docs/NEWS.md
@@ -2,6 +2,7 @@
 
 Full news history for ClawCodex. The [README News section](../README.md#-news) keeps only the 10 most recent items.
 
+- **2026-07-29 (v1.3.0):** **ClawCodex scores 80.9% on Terminal-Bench 2.1 — a top-tier open-source result on Opus 5 (#720–#725, #747–#754)** — running headless on `claude-opus-5` at `effort=xhigh`, ClawCodex solved **72 of 89** Terminal-Bench 2.1 tasks: **80.9% pass@1** on a single run. On the [public 2.1 leaderboard](https://www.tbench.ai/leaderboard/terminal-bench/2.1) (k=5 averages) that would slot **around third** — behind Claude Code / Fable 5 (83.8%) and Codex / GPT-5.5 (83.1%), statistically level with the 79–80% cluster, and **ahead of Claude Code on Opus 4.8 (78.9%) and Sonnet 5 (74.6%)**. Getting there was open, unglamorous parity work: a Harbor eval adapter (`eval/harbor/`) for three-way ClawCodex-vs-openclaude-vs-Claude-Code runs (#720, #724, #725), then a run of prompt- and reliability-parity fixes — restored task-tool skip conditions and parallel-tool guidance, deferred nonessential initial tools, and recovery of trials lost to empty turns and transport drops (#747–#754). **Also in v1.3.0:** `claude-opus-5` support with an interactive `/effort` fix (#746), bounded persistent memory with a background self-improvement review (#731), a VS Code extension driving the agent-server over stdio (#727), image-paste input with an `[Image #N]` un-attach chip (#761, #762), the `CLAUDE.md → CLAWCODEX.md` context-file rebrand (#732), and transport-retry hardening (#757, #760). Stated plainly: this is a single k=1 pass (binomial 1σ ±4.2pp) against the board's k=5 ± ~1.2pp averages, benchmarked on `main` at #756 (before the v1.3.0 tag), so read it as directional rather than a ranked submission.
 - **2026-07-13:** **`/eco` token compression — -80% Bash-output tokens, measured, now a headline (#708, #712)** — a new session toggle compresses the model-bound rendering of every Bash result with deterministic filters ported from [RTK](https://github.com/rtk-ai/rtk)'s method set: failure-focused test summaries (kept error lines are never rewritten), `git`/`pip`/`npm` ceremony stripping, log dedup with `[×N]` counts, and a recoverable head-cap — all behind a **never-worse** guard, with every lossy compression teeing the full output to disk behind a runnable recovery hint (#708). A reproducible benchmark (`eval/eco/`) replays 27 real command outputs through the exact production pipeline and counts real tokenizer tokens: **92,989 → 17,767 (-80%)** corpus-wide, -88% on filter hits, plus an honestly conservative recompute of RTK's own 30-minute-session model (-19% under their averaged assumptions — real sessions are fat-tailed) (#712). Full tables: the [`/eco` section](../README.md#eco-benchmark) and [`eval/eco/results/`](../eval/eco/results/results.md).
 - **2026-07-12 (v1.1.0):** **ClawCodex v1.1.0 — run OpenAI *and* Claude models on your subscription, not metered API billing** — the headline of 1.1.0 is **subscription auth for the two biggest model families**, so you can point ClawCodex at a plan you already pay for. **Sign in with ChatGPT (#698):** `clawcodex login → openai → subscription` (browser, device-code, or import from an existing Codex CLI login) routes requests through the ChatGPT Codex backend's Responses API — `gpt-5.5`, `gpt-5.4`, `gpt-5.4-mini`, and `gpt-5.3-codex-spark` on your Plus/Pro allowance, with encrypted-reasoning replay across turns and **$0** metered cost. **Claude Pro/Max (#697):** `clawcodex login → anthropic → subscription` connects a Claude subscription via OAuth (PKCE) with automatic token refresh and the same $0 accounting; follow-ups repaired the login after Anthropic moved its OAuth endpoints to `platform.claude.com` (#702) and stopped sending adaptive thinking to models that don't support it (#699). A configured API key always wins, and subscription usage reports `billing_mode: subscription`. **More models:** a Meta (`api.meta.ai`) provider with the 1M-context `muse-spark-1.1` reasoning model (#692) and refreshed MiniMax parameters (#696). **Workflow & TUI:** `/plan` mode with implicit plan-mode entry/exit (#676), `--worktree/-w` session isolation for parallel runs in separate git worktrees (#672), the `/memory` picker + `$EDITOR` spawn (#693), config/state directories rebranded `.claude → .clawcodex` with a one-time migration (#678), `/logo` startup color schemes (#677), plus TUI polish — Tab accepts the suggested placeholder (#690), past inputs get the Claude-Code highlight band (#691), clickable agent URLs (#694), and a per-terminal link-open affordance (#701). **Quality:** semantic tool-input coercion with parity validation errors (#700) and looser, Claude-Code-faithful permission granting (#673).
 - **2026-07-07:** **`/loop` scheduled tasks now actually fire — full port of Claude Code's session-scoped scheduler (#680)** — the bundled `/loop` skill finally has a real engine behind it: a new `src/scheduled_tasks` module parses standard 5-field cron expressions and fires due prompts **between turns** from the agent-server's idle poll. `CronCreate`/`CronList`/`CronDelete` register real firing jobs (8-char IDs, 50-job cap, deterministic jitter, 7-day recurring expiry with one final fire), and the new **`ScheduleWakeup`** tool drives self-paced `/loop` mode — the model picks each next delay (1 min–1 hr), `stop: true` ends the loop, and a ~20-minute fallback wakeup catches iterations that forget to reschedule. Typed skill slash commands now reach the backend (new `skill_command` control), so `/loop 5m check ci` works from the composer with completion + argument hint; the TUI shows a live countdown indicator (`⟳ loop wakeup in 2m 14s · ⏰ 1 scheduled`) and **Esc while idle stops a waiting loop**. `/clear` drops session tasks, `--resume` restores unexpired ones, `CLAWCODEX_DISABLE_CRON=1` disables the scheduler. 117 new tests; verified live over stdio NDJSON and a real PTY TUI drive (typed dispatch → CronCreate → a real wakeup fire between turns → Esc-stop).