RepoJournal
OpenClaw @openclaw
159 wires and counting
$ status: archived — this briefing has stopped; the archive stays readable. browse live briefings →

$ status

wire 2026-06-26
stories 62
GitHub RSS Archive Reviews

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Exec workdir checks fail early, Telegram reply loops capped

By RepoJournal · Filed · About OpenClaw · Composed from the cited sources · methodology

Renaud Cerrato obviyus ZengWen-DT VACInc vincentkoc openclaw-docs-sync[bot]
6 people shipped this
Renaud Cerrato @renaudcerrato 2 cited obviyus @obviyus 1 cited ZengWen-DT @ZengWen-DT 1 cited VACInc @VACInc 1 cited vincentkoc @vincentkoc openclaw-docs-sync[bot] @openclaw-docs-sync[bot]

OpenClaw exec now rejects invalid explicit workdirs before running commands, and Telegram reply-session initialization is serialized to stop replay loops.

OpenClaw exec now fails an explicit local `exec.workdir` that is unavailable instead of silently falling back to another directory. The command no longer runs from the wrong location, and a regression test verifies that behavior. Implicit or default workdirs still retain their fallback behavior [1][2].

Telegram isolated-ingress replayed the same spooled update forever when reply-session initialization hit a repeated stale-snapshot conflict. The stale-snapshot guard read outside the session-store writer queue, so multiple same-session initializers could compute from stale data. Initialization is now serialized through the queue, ending the retryable conflict loop [3]. A follow-up backs off retryable Telegram spool replays that fail with `reply session initialization conflicted...` so one pending update cannot hot-loop the lane [4].

The chat-history truncation log underreported dropped messages: `enforceChatHistoryFinalBudget` returned `placeholderCount: 0` when keeping only the last message, silently dropping N-1 earlier ones. The caller only logs truncation when that count is positive. The report now reflects omitted messages [5].

$ ls openclaw/reviews/ # the step-back reads

the week in review Openclaw 2026.8.35 extended-stable ships critical security fixes the month in review OpenClaw 2026.9.7 backs up state before migrations
Full archive → Weekly reviews →

Quick answers

What shipped in OpenClaw on June 26, 2026?
OpenClaw exec now rejects invalid explicit workdirs before running commands, and Telegram reply-session initialization is serialized to stop replay loops. In total, 36 commits and 26 pull requests landed.
Who contributed to OpenClaw on June 26, 2026?
6 developers shipped this update, including Renaud Cerrato, obviyus, ZengWen-DT, VACInc, vincentkoc, and openclaw-docs-sync[bot].
What were the notable OpenClaw updates?
fix(exec): fail invalid explicit workdir before running (#94441), fix(exec): fail invalid explicit workdir before running, and fix(auto-reply): serialize reply session initialization.
♥

$ log @openclaw --recent

ChatGPT-login model picker re-pinned to bundled Codex 221 shipped OpenClaw fixes premature approval expiry, embedding readiness, and startup integrity proofs 148 shipped Gateway PR comparisons go local, auto-reply compaction stops eating turns 122 shipped Startup admission errors typed, setup diagnostics unhidden 166 shipped

$ log --all # elsewhere on the wire

Agent Security Quiet day in agent security as guard coverage reaches four clients Alignment & Post-training verl defaults Ascend RL mode, open-instruct unblocks SFT packing

$ ls showcases/

Safety Evals AI Red Teaming Agent Security Interpretability Alignment & Post-training

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.