Clean up SDK shell tool failure modes

Three concrete wraps on exec_command / write_stdin via the existing
Shell capability configure_tools mechanism, plus one skill-doc fix.
All wraps fire on both Responses and chat-completions paths; the
chat-completions error-as-result wrap still stacks on top when needed.

- write_stdin: decode the common escape forms in `chars` (\uXXXX,
  \xXX, \n \t \r \0 \a \b \v \f \\). Models routinely send the
  literal six-char string `` intending the ASCII control byte;
  the SDK takes chars verbatim so the byte never reaches the PTY and
  documented mechanisms like Ctrl-C, arrows, and Escape silently
  don't work. Allowlist regex over recognized escapes only —
  unrecognized sequences like `\p` pass through untouched.

- exec_command: catch InvalidManifestPathError and rewrite to a
  model-actionable message ("workdir must be a path inside
  /workspace") using the exception's structured `context["rel"]` so
  we don't need to string-match the SDK's wording.

- Both tools: catch pydantic ValidationError once at the wrap and
  reformat into a short "{tool}: invalid arguments — {field}: {msg}"
  string. Covers empty cmd, missing required fields, ge/min_length
  violations on max_output_tokens and yield_time_ms — and any future
  schema field the SDK adds.

Updated python.md guidance: the `shell=` parameter is for swapping
POSIX shells (bash/zsh/sh). Interpreters belong in `cmd` —
`cmd="python3 -c '...'"`, not `shell=python3`. The `shell=interpreter`
shortcut breaks in interpreter-specific ways (python needs `-c`,
node/ruby/perl need `-e`) so there's no clean code fix and we don't
try one.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
0xallam
2026-05-25 14:05:42 -07:00
parent fdd8b71f4e
commit 418eedcd41
2 changed files with 109 additions and 5 deletions
+101 -5
View File
@@ -20,12 +20,15 @@ from __future__ import annotations
import inspect
import json
import logging
import re
from typing import TYPE_CHECKING, Any
from agents.agent import ToolsToFinalOutputResult
from agents.sandbox import SandboxAgent
from agents.sandbox.capabilities import Filesystem, Shell
from agents.sandbox.errors import InvalidManifestPathError
from agents.tool import CustomTool, FunctionTool, Tool
from pydantic import ValidationError
from strix.agents.prompt import render_system_prompt
from strix.tools.agents_graph.tools import (
@@ -180,10 +183,103 @@ def _configure_chat_completions_filesystem_tools(toolset: Any) -> None:
setattr(toolset, name, _function_tool_with_error_result(tool))
def _configure_chat_completions_shell_tools(toolset: Any) -> None:
_CHARS_ESCAPE_RE = re.compile(r"\\(?:u[0-9a-fA-F]{4}|x[0-9a-fA-F]{2}|[0abtnvfr\\])")
_CHARS_ESCAPE_MAP = {
"\\\\": "\\",
"\\n": "\n",
"\\t": "\t",
"\\r": "\r",
"\\0": "\x00",
"\\a": "\x07",
"\\b": "\x08",
"\\v": "\x0b",
"\\f": "\x0c",
}
def _decode_chars_escape(s: str) -> str:
if "\\" not in s:
return s
def sub(match: re.Match[str]) -> str:
token = match.group(0)
if token in _CHARS_ESCAPE_MAP:
return _CHARS_ESCAPE_MAP[token]
if token.startswith(("\\u", "\\x")):
return chr(int(token[2:], 16))
return token
return _CHARS_ESCAPE_RE.sub(sub, s)
def _format_validation_error(tool_name: str, exc: ValidationError) -> str:
parts: list[str] = []
for err in exc.errors():
loc = ".".join(str(x) for x in err.get("loc", ()))
msg = err.get("msg", "invalid")
parts.append(f"{loc}: {msg}" if loc else msg)
return f"{tool_name}: invalid arguments — " + "; ".join(parts)
def _wrap_exec_command(tool: FunctionTool) -> FunctionTool:
invoke_tool = tool.on_invoke_tool
async def invoke(ctx: Any, raw_input: str) -> Any:
try:
return await invoke_tool(ctx, raw_input)
except ValidationError as exc:
return _format_validation_error(tool.name, exc)
except InvalidManifestPathError as exc:
rel = exc.context.get("rel", "?")
return (
"exec_command: workdir must be a path inside /workspace "
"(or omitted to use the turn's cwd). "
f"Got: {rel!r}."
)
tool.on_invoke_tool = invoke
return tool
def _wrap_write_stdin(tool: FunctionTool) -> FunctionTool:
invoke_tool = tool.on_invoke_tool
async def invoke(ctx: Any, raw_input: str) -> Any:
try:
parsed = json.loads(raw_input)
except json.JSONDecodeError:
parsed = None
if isinstance(parsed, dict) and isinstance(parsed.get("chars"), str):
parsed["chars"] = _decode_chars_escape(parsed["chars"])
raw_input = json.dumps(parsed)
try:
return await invoke_tool(ctx, raw_input)
except ValidationError as exc:
return _format_validation_error(tool.name, exc)
tool.on_invoke_tool = invoke
return tool
def _configure_shell_tools(toolset: Any, *, chat_completions: bool) -> None:
for name, tool in vars(toolset).items():
if isinstance(tool, FunctionTool):
setattr(toolset, name, _function_tool_with_error_result(tool))
if not isinstance(tool, FunctionTool):
continue
wrapped = tool
if tool.name == "exec_command":
wrapped = _wrap_exec_command(wrapped)
elif tool.name == "write_stdin":
wrapped = _wrap_write_stdin(wrapped)
if chat_completions:
wrapped = _function_tool_with_error_result(wrapped)
setattr(toolset, name, wrapped)
def _make_shell_configurator(*, chat_completions: bool) -> Any:
def configure(toolset: Any) -> None:
_configure_shell_tools(toolset, chat_completions=chat_completions)
return configure
def _lifecycle_tool_completed(tool_name: str, output: Any) -> bool:
@@ -366,8 +462,8 @@ def build_strix_agent(
),
),
Shell(
configure_tools=(
_configure_chat_completions_shell_tools if chat_completions_tools else None
configure_tools=_make_shell_configurator(
chat_completions=chat_completions_tools,
),
),
],
+8
View File
@@ -11,6 +11,14 @@ Prefer writing reusable scripts to `/workspace/scratch/<name>.py` and
running them with `python3 /workspace/scratch/<name>.py`. For short
one-off transformations, `python3 -c` or a small here-document is fine.
The `shell` parameter on `exec_command` is for swapping POSIX shells
(`bash`/`zsh`/`sh`), not for picking interpreters. Put the interpreter
invocation in `cmd` instead: `cmd="python3 -c '...'"`, not
`shell=python3, cmd="..."`. The `shell=<interpreter>` shortcut breaks
in subtle ways — `python3` works only with `login=False` (because the
SDK adds `-l`/`-i`), and other interpreters (`node`, `ruby`, `perl`)
take `-e` not `-c` so they fail even with `login=False`.
## Proxy Automation From Python
The sandbox image includes an installed `caido_api` module. Import it