SWE-agent · Issues· 113 open

Chooser swallows cost-limit exceptions (bare except Exception) and crashes with UnboundLocalError, defeating the documented abort

BinaryTrajectoryComparisonConfig.comparison_temperature is silently discarded: the comparison query runs at the model's own temperature

[First-time contributor] Looking for beginner-friendly issues to contribute

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

[Bug]: windowed edit tool parses replace_all with type=bool, so "false"/"0" enable replace-all

feat(tools): add optional CodeGraph tool bundle

Submission integrity: outcome decided in payload space; gold test patch readable; PS1 sentinel spoof; failed instances vanish from preds; blocklist is first-token-only

Path traversal in the trajectory inspector reads off-path trajectory/JSON files (unauthenticated, all-interfaces bind, wildcard CORS)

[Bug]: ScoreRetryLoop treats zero API calls as infinity when breaking score ties

Verified OSS loop: claim leases, evidence receipts, and independent exact-head review

Idea: EvalPort ResultSet as a portable export for compare_runs.py-style comparisons

RFC: opt-in deterministic guard for shell commands inside the Docker sandbox (destructive + secret-exfil, <2ms)

Checkout fails on default Git-for-Windows: tracked paths up to 180 chars overflow MAX_PATH

sweagent --help crashes with UnicodeEncodeError on default Windows consoles (cp1252 stdio)

BUG: BinaryTrajectoryComparison skips extra sampling for indented edit actions