All evidence cases
DeepSWE v1.1repositoryPython

Add interprocedural taint checks for Bandit injection sinks

Track user input through variables and calls so Bandit flags injection sinks beyond string literals.

bandit-interprocedural-taint-checks

Normalized profile

What this task asks for

Track user input through variables and calls so Bandit flags injection sinks beyond string literals. Track input taint from Flask request data, `sys.argv`, `input()`, and `os.environ` into security-sensitive sinks. Propagate taint through string construction, calls, multi-hop assignments, nested functions, and import aliases. Treat parameterized query values, numeric conversion, path basenames, shell quoting, and HTML escaping as safe. Add B620 through B624 checks for SQL injection, shell injection, path traversal, SSRF, and XSS sinks. Report every new plugin finding with HIGH severity and MEDIUM confidence.

  • feature implementation
  • security remediation
  • data transformation
  • query layer
  • serialization
  • Bandit
  • SQL
  • repository patch
  • security-sensitive or adversarial behavior
  • specialized algorithmic or domain reasoning

Published evidence

Model × harness outcomes

Counts are tied to these exact configurations. Harnesses are not treated as equivalent across benchmarks.

Models
13
Configurations
41
Trials
164

Showing 41 of 41 configurations

ModelHarnessEffortResultPasses
claude-fable-5vertex_aimini-swe-agentpublished configurationlowMixed3/4
claude-fable-5vertex_aimini-swe-agentpublished configurationmediumMixed2/4
claude-fable-5vertex_aimini-swe-agentpublished configurationhighAll passed4/4
claude-fable-5vertex_aimini-swe-agentpublished configurationxhighAll passed4/4
claude-fable-5vertex_aimini-swe-agentpublished configurationmaxMixed1/4
claude-opus-4-8anthropicmini-swe-agentpublished configurationlowMixed2/4
claude-opus-4-8anthropicmini-swe-agentpublished configurationmediumMixed2/4
claude-opus-4-8anthropicmini-swe-agentpublished configurationhighMixed2/4
claude-opus-4-8anthropicmini-swe-agentpublished configurationxhighMixed3/4
claude-opus-4-8anthropicmini-swe-agentpublished configurationmaxAll passed4/4
claude-sonnet-4-6anthropicmini-swe-agentpublished configurationhighNo passes0/4
claude-sonnet-5anthropicmini-swe-agentpublished configurationlowMixed3/4
claude-sonnet-5anthropicmini-swe-agentpublished configurationmediumMixed2/4
claude-sonnet-5anthropicmini-swe-agentpublished configurationhighAll passed4/4
claude-sonnet-5anthropicmini-swe-agentpublished configurationxhighAll passed4/4
claude-sonnet-5anthropicmini-swe-agentpublished configurationmaxAll passed4/4
gemini-3-1-pro-previewvertex_aimini-swe-agentpublished configurationhighNo passes0/4
gemini-3-5-flashvertex_aimini-swe-agentpublished configurationmediumMixed1/4
glm-5-2zaimini-swe-agentpublished configurationhighMixed2/4
glm-5-2zaimini-swe-agentpublished configurationmaxMixed2/4
gpt-5-4openaimini-swe-agentpublished configurationxhighMixed3/4
gpt-5-5openaimini-swe-agentpublished configurationlowNo passes0/4
gpt-5-5openaimini-swe-agentpublished configurationmediumNo passes0/4
gpt-5-5openaimini-swe-agentpublished configurationhighMixed2/4
gpt-5-5openaimini-swe-agentpublished configurationxhighMixed2/4
gpt-5-6-lunaopenaimini-swe-agentpublished configurationlowNo passes0/4
gpt-5-6-lunaopenaimini-swe-agentpublished configurationmediumNo passes0/4
gpt-5-6-lunaopenaimini-swe-agentpublished configurationhighMixed1/4
gpt-5-6-lunaopenaimini-swe-agentpublished configurationxhighMixed3/4
gpt-5-6-lunaopenaimini-swe-agentpublished configurationmaxAll passed4/4
gpt-5-6-solopenaimini-swe-agentpublished configurationlowMixed3/4
gpt-5-6-solopenaimini-swe-agentpublished configurationmediumMixed2/4
gpt-5-6-solopenaimini-swe-agentpublished configurationhighAll passed4/4
gpt-5-6-solopenaimini-swe-agentpublished configurationxhighAll passed4/4
gpt-5-6-solopenaimini-swe-agentpublished configurationmaxAll passed4/4
gpt-5-6-terraopenaimini-swe-agentpublished configurationlowNo passes0/4
gpt-5-6-terraopenaimini-swe-agentpublished configurationmediumMixed1/4
gpt-5-6-terraopenaimini-swe-agentpublished configurationhighMixed1/4
gpt-5-6-terraopenaimini-swe-agentpublished configurationxhighNo passes0/4
gpt-5-6-terraopenaimini-swe-agentpublished configurationmaxMixed3/4
kimi-k2-7-codeopenroutermini-swe-agentpublished configurationdefaultNo passes0/4