* fix(agents): edit tool rewrites line endings on lines it did not touch
The edit tool detected one line ending from the file's first newline,
normalized the whole file to LF for matching, then re-applied that single
ending to every line on write. A one-line edit therefore rewrote the
terminator of every untouched line, and deleted lone carriage returns
that were data rather than line breaks. The diff and unified patch
returned to the model are computed on the normalized text, so the damage
never surfaces in the tool result or the TUI preview.
Rebuild the written text from the original terminators instead. A bare
carriage return is tracked as its own terminator, so lines the edits did
not touch keep their exact bytes and a rewritten line is written back
with the terminator bytes it had. Newly written lines take the
terminator of the original line they replaced. Matching still runs in LF
space, so LF oldText against a CRLF file matches as before.
* fix(agents): preserve CR fallback for leading edit insertions
* fix(agents): align replacement line-ending boundaries
* fix(agents): preserve edit line-ending provenance
* refactor(agents): isolate edit replacement reconstruction
* test(agents): cover edits through production line endings
---------
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
The claude-cli terminal result envelope carries only the final message text.
On the default draft-preview path (no block streaming, no commentary
classification) the preview streamed every text block, then final delivery
edited it down to just the post-tool closer, silently erasing pre-tool
answer text (#106760).
Both the incremental parser and the transcript reparse now keep the
streamed accumulation when it is connected to the result through tool
splits: boundaries join with a paragraph break, non-tool message
boundaries restart preservation candidacy (superseded drafts still defer
to the envelope), interim results commit their segment without breaking
cumulative delta snapshots, and transcript reparses continue past interim
results like the incremental parser.
Fail build and copy generators with ETIMEDOUT after 600000ms and terminate the complete managed process tree. Keep the timeout policy in the bundled-assets caller while preserving the shared helper for distinct callers such as #111349.
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
The Diffs plugin named the `message` tool from three model-visible surfaces:
the guidance it injects via before_prompt_build, the mode=file tool result,
and its SKILL.md. `disableMessageTool` removes that tool in eight production
paths (subagent spawn, system agent, worker runtime, both skills-workshop
reviews, companion-ask, restart sentinel, active-memory recall), so agents in
those sessions were told to call a tool absent from their tool list.
The same surfaces plus the `diffs` tool description named the separately
gated Canvas plugin — SKILL.md through `canvas present` / `canvas navigate`,
the others through "for canvas use".
All of them now describe the capability instead of naming the tool, per the
root AGENTS.md rule that descriptions never statically name tools from other
toolsets. The two pinned assertions keep their snapshots and now also assert
the invariant directly.
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
* feat(android): add media rendition and video support
* chore(android): refresh native i18n inventory
* fix(android): opt in to Media3 command API
* chore(android): sync native i18n source locations
* feat(cli): run agent exec against the ambient config, composed in memory
Exec previously ignored the operator's config entirely, so a one-shot turn
could not reach configured providers, credentials, or agentRuntime harness
selection. It now layers config the way other folder-scoped coding CLIs do.
The composed config is published as this process's runtime snapshot rather
than serialized to a temp file and re-read through OPENCLAW_CONFIG_PATH. The
snapshot is the only in-process config cache, so the file only ever fed it --
while writing env-substituted provider keys to disk where the run's own exec
tool could read them.
* fix(cli): resolve exec stored credentials from the configured agent dir
* chore(scripts): allow agent exec the file-scoped config loader at its process boundary
* test(cli): cover the exec credential default and pinned-config flags
* fix: diagnose prepared model owner config drift
* fix(gateway): bind replies to published model owner
* fix(gateway): bind durable replay to published owner
* test: preserve requested workspace in reply mock
