emrg: host scripts must not print output a legacy console codec cannot encode

Agent.md

@@ -119,7 +119,7 @@ Community needs voiced in HN agent-UI discussions map directly to EMRG's design:
 pkill -f "emrg.server"; rm -f ~/.emrg/emrgd.token; python -m emrg
 ```
 
-Python: `uv run pytest tests/ -v` (1277) — import check: `uv run python -c "from emrg.client.app import run_client"`
+Python: `uv run pytest tests/ -v` (1284) — import check: `uv run python -c "from emrg.client.app import run_client"`
 GUI: `cd emrg/gui && npm test` (100: 44 daemon_client + 20 conn-manager + 7 integration + 7 nav-policy + 7 gui-state + 6 build-config + 4 boot-contract + 3 preload-api + 2 theme-guard) — syntax: `node --check main.js preload.js daemon_client.js`
 Renderer: `cd emrg/gui/renderer && npm run typecheck && npm test` (514: 5 snapshot-store + 9 utils + 3 ErrorBoundary + 2 App smoke + 11 commands + 4 copywriting + 11 i18n + 13 markdown + 21 transcript + 11 TranscriptView + 15 history + 31 composer + 41 Composer + 6 LinkDialog + 16 sidebar + 17 Sidebar + 9 fileTree + 9 FileTree + 16 resultPanel + 8 ResultPanel + 27 workspaceView + 10 WorkspaceView + 10 dialog + 6 Dialog + 9 ConfirmDialog + 9 RenameDialog + 10 dialogLists + 3 HelpDialog + 9 MemoryDialog + 6 SkillsDialog + 8 openSession + 6 WelcomeDialog + 9 OpenSessionDialog + 7 NewSessionDialog + 7 rewind + 8 RewindDialog + 7 GithubDeviceDialog + 18 daemonBridge + 7 DaemonBridgeProvider + 30 Shell + 15 DialogHost + 20 SettingsPanel + 6 TaskFormDialog + 5 RantDialog + 4 vendorMarkdown) + `npm run build` → `renderer/dist/`
 CI: `uv run pytest` (ubuntu + **windows-2025 matrix** — Windows pytest 回归在 PR CI 即失败，v0.2.29 教训 #725) + GUI tests + **actionlint workflow lint** (`rhysd/actionlint@v1.7.12` gate, #444 — workflow 解析错误在 PR CI 即失败，如 `if:` secrets 上下文)

scripts/calibrate_silent_drift_threshold.py

@@ -275,7 +275,7 @@ def main(argv: list[str] | None = None) -> int:
         print(
             f"\nNo sub-threshold observations accumulated yet "
             f"(n_noise=0, n_drift={stats['n_drift']}). The detector writes an "
-            f"anchor_bias_observation per anchored round — re-run after the "
+            f"anchor_bias_observation per anchored round - re-run after the "
             f"daemon has seen some usage. Current threshold stays "
             f"{args.current}."
         )
@@ -316,15 +316,15 @@ def main(argv: list[str] | None = None) -> int:
     if reason == "noise-crowding-boundary":
         print(
             f"\nNOISE CROWDING THE BOUNDARY: p99 ({_fmt(stats['p99'])}) is within "
-            f"10% of the current threshold {args.current} — per-round estimate "
+            f"10% of the current threshold {args.current} - per-round estimate "
             f"noise is close to tripping the guard. Recommended "
             f"_SILENT_DRIFT_THRESHOLD: {stats['recommended']} "
             f"(= max(current, p99 * 1.5))."
         )
     elif reason == "noise-well-below-boundary":
         print(
             f"\nNoise is well below the boundary (p99={_fmt(stats['p99'])}, "
-            f"current={args.current}) — the guard demonstrably fires only on "
+            f"current={args.current}) - the guard demonstrably fires only on "
             f"real drift. Recommended _SILENT_DRIFT_THRESHOLD: "
             f"{stats['recommended']} (unchanged)."
         )
@@ -333,7 +333,7 @@ def main(argv: list[str] | None = None) -> int:
             f"\nNO CLEAN SEPARATION: noise tail p99 ({_fmt(stats['p99'])}) * 1.5 "
             f"exceeds the smallest observed drift "
             f"({_fmt(stats['min_drift']) if stats['min_drift'] is not None else 'n/a'} "
-            f"* 0.75 headroom) — a single fixed threshold cannot separate the "
+            f"* 0.75 headroom) - a single fixed threshold cannot separate the "
             f"two populations. Keeping current threshold {args.current}; "
             f"consider per-provider thresholds or investigating the noisy "
             f"providers."

scripts/check_nonlocal.py

@@ -175,7 +175,7 @@ def check_nonlocal(app_path: str) -> int:
             exit_code = 1
 
     if exit_code == 0:
-        print("✅ nonlocal integrity check passed")
+        print("OK: nonlocal integrity check passed")
 
     return exit_code
 

scripts/push-branch-from-api.py

@@ -4,20 +4,20 @@
 Counterpart of scripts/sync-master-from-api.py (fetch direction). EMRG repeatedly
 hits github.com:443 unreachable while api.github.com stays up (10+ documented
 cycles, 08-22..08-26). When a fix must be pushed during an outage, the old flow
-was a hand-written ephemeral script re-derived each time from memory notes —
+was a hand-written ephemeral script re-derived each time from memory notes -
 cycle 2026-08-26 00:59 recorded four gotchas learned the hard way:
 
   1. blobs must be uploaded from the *committed* object bytes (`git cat-file
      blob <sha>`), not working-tree bytes (CRLF normalization differs)
   2. `git ls-tree` needs `-r` to enumerate nested paths
-  3. the API displays dates as UTC ('Z') but stores the raw offset (+0800) —
+  3. the API displays dates as UTC ('Z') but stores the raw offset (+0800) -
      recreate commits with the original epoch+offset or the sha will not match
   4. subprocess text I/O must use encoding='utf-8', errors='replace' (GBK
      console crashes on non-ASCII commit messages)
 
 This script automates that recipe:
 
-  * walks the local chain from <ref> (default HEAD) down to the remote base —
+  * walks the local chain from <ref> (default HEAD) down to the remote base -
     the existing branch head, or the first ancestor the remote already has
     (GET /repos/{repo}/commits/{sha})
   * uploads blobs (raw bytes, byte-exact) and trees (structured entries,
@@ -272,7 +272,7 @@ def push_branch(repo: str, branch: str, ref: str, force: bool, cwd: str | None =
     if not chain:
         raise PushError("internal: empty commit chain")
 
-    print(f"  base: {base or '(new branch — nothing on remote yet)'}")
+    print(f"  base: {base or '(new branch - nothing on remote yet)'}")
     print(f"  uploading {len(chain)} commit(s)...", flush=True)
 
     # ---- upload objects bottom-up, mapping local -> remote shas
@@ -428,7 +428,7 @@ def main() -> int:
     print(f"repo {repo} branch {args.branch}")
     print(f"  remote sha: {result['sha']}")
     if result["result"] == "no-op":
-        print("  nothing to do — ref already at that commit")
+        print("  nothing to do - ref already at that commit")
     else:
         print(f"  original local tip: {result['original_local_tip']}")
         print(f"  content identical to local tip: {result['content_identical']}")

scripts/reader_fix_latency.py

@@ -2,7 +2,7 @@
 """Measure the reader-feedback -> merged-fix loop: median issue-to-fix latency.
 
 Issue #1027 secondary suggestion (reader comment heinrichneb, Dev.to 3dicj):
-"add a README metric — median time from reader-found boundary to merged fix
+"add a README metric - median time from reader-found boundary to merged fix
 (the 50-min loop claim becomes a measured number)."
 
 The Dev.to article claimed a ~50-minute issue-to-merged-fix loop (#1000: issue
@@ -16,10 +16,10 @@
     issue.created_at  ->  merged_at of the EARLIEST merged linked PR
 
 The median across all such issues is the headline number. Only the earliest
-merged PR counts — a fix is "shipped" when its first merged PR lands.
+merged PR counts - a fix is "shipped" when its first merged PR lands.
 
 The script also reports the OPEN side of the loop (issue #1056, Dev.to
-comment 3djoa, heinrichneb 2026-08-27): an open-issue age counter — how
+comment 3djoa, heinrichneb 2026-08-27): an open-issue age counter - how
 many open issues exist, their median age in days, and the oldest few. A
 median only over already-fixed issues hides issues that are still waiting;
 the two numbers together bound the loop from both ends.
@@ -180,7 +180,7 @@ def main(argv: list[str] | None = None) -> int:
 
     samples = collect_latencies(args.repo, args.limit)
     if not samples:
-        print("no closed issues with a merged fix PR found — nothing to measure")
+        print("no closed issues with a merged fix PR found - nothing to measure")
         return 0
 
     vals = [s[1] for s in samples]

scripts/sync-master-from-api.py

@@ -5,12 +5,12 @@
 stays up (10+ documented cycles, e.g. 08-22..08-26). The usual fallback flow:
 the local repo already contains the content (a branch pushed via the Git Data
 API that later got squash-merged upstream), so advancing local refs only needs
-the missing *commit* objects — reconstructed byte-exact from the API's
+the missing *commit* objects - reconstructed byte-exact from the API's
 verification payload + signature, including web-flow GPG-signed squash merges.
 
 
 When the head commit's *content* objects (blobs/trees) are also missing locally
-(e.g. a parallel PR introduced files this repo never had — first hit in cycle
+(e.g. a parallel PR introduced files this repo never had - first hit in cycle
 cyc20260826-154904 with #994's GUI assets), the script now auto-fetches them via
 the Git Data API: blobs via `git/blobs/{sha}` + `git hash-object -w`, trees via
 `git/trees/{sha}` + `git mktree` (canonical ordering), recursing bottom-up. The
@@ -27,7 +27,7 @@
     `git hash-object -t commit -w` (byte-exact, GPG signature preserved)
   * verifies the root tree sha matches the remote; if content objects are
     missing, fetches missing blobs/trees via the Git Data API (disable with
-    --no-fetch-objects) and re-verifies — fail-loud only if still mismatched
+    --no-fetch-objects) and re-verifies - fail-loud only if still mismatched
   * updates refs/heads/<ref> and refs/remotes/origin/<ref>
 
 Requirements: git on PATH; api.github.com reachable. Auth: optional for public
@@ -228,7 +228,7 @@ def main() -> int:
                                 capture_output=True, text=True)
     if local_tree.returncode != 0 or local_tree.stdout.strip() != tree:
         if not args.no_fetch_objects:
-            print(f"  root tree {tree[:7]} missing locally — fetching blobs/trees via Git Data API")
+            print(f"  root tree {tree[:7]} missing locally - fetching blobs/trees via Git Data API")
             _fetch_tree(repo, tree)
             local_tree = subprocess.run(["git", "rev-parse", head + "^{tree}"],
                                         capture_output=True, text=True)