Repository Issue Activity (beta)

#953 RFC: deterministic execution boundaries for tool calls (task text is untrusted input)

#916 Your project is on StackMap — a curated map of the AI stack

#949 Question: hosted OpenAI-compatible api_base example (not localhost)

#921 Permanent context-overflow (400) retried 10 times — full oversized history resent each attempt

#950 [ Important ! ]The default mode causes meaningless sleep

#947 Emoji rendering issues

#945 Try to reduce context when the LLM response fails

#938 # Successful submissions are discarded when the shell writes startup output to stdout

#941 Pin litellm / ship a lock file — floating "litellm >= 1.75.5" causes intermittent runtime breakage

#937 Test suite cannot run on Windows: cmd.exe quoting desyncs the scripted agents (144 failures)

#909 DefaultAgent.run() retains per-run budget and timing state across invocations

#931 mini-extra config edit fails on Windows when EDITOR is unset

#927 Question: Is MCP support planned for mini-swe-agent?

#897 Why is the model's malformed response dropped from history on format-error retry?

#914 Format errors are billed but not counted, so the agent runs past cost_limit

#895 feat(env): Add support for configurable shell (pwsh) in LocalEnvironment

#907 How do i run mini-swe-agent in an air gapped environment?

#889 LitellmModel._query hardcodes tools=[BASH_TOOL] — no way to add tools without copying the whole method

#899 The default model class uses the wrong `format_error_template`

#898 Withdrawn: opened in wrong repository

#894 Malformed tool-call format errors are masked by the "output truncated" template branch (finish_reason == "tool_calls")

#893 Bug: _run leaks file descriptor and loses partial output when os.killpg raises ProcessLookupError

#885 `import minisweagent` crashes on non-UTF stdout (Windows cp1252)

#874 mini-swe-agent hangs while the provider stalls mid tool-call stream

#873 mini-swe-agent forwards large shell output back to the provider