Tool · Codex · OpenAI

exec_command

How this differs from most implementations
  • Binary content never enters context silentlyinvariantfailsvs holds in 4 of 4
  • Output wrapperraw stdoutvs plain in 6 of 8
  • Default line capnonevs 2000 in 6 of 7
  • Byte / char capnonevs 50 KB in 3 of 5
  • Per-line capnonevs 2000 chars in 4 of 4
  • Giant single linenonevs per-line cap in 4 of 6

Derived by comparing every recorded value against the other implementations of the same capability. Differing is not a defect unless the row is marked an invariant.

Reached indirectly. Reached through the shell rather than a dedicated tool.

Observed · Read

compare all →
Output shape
Dedicated read toolnone — cat/sed via exec_commandsrc
Line numberingnonesrc
Raw bytes, no numbering at all.
Output wrapperraw stdoutsrc
Limits
Default line capnonesrc
Byte / char capnonesrc
Per-line capnonesrc
Whichever hits first
Pagination
Offset base
n/a
Announces truncationinvariant
No read tool; the shell returns whatever the command produced.
Continuation hint
Reports total lines
Out-of-range offset
n/a
Non-text
Binary content never enters context silentlyinvariantfailssrc
Reading through the shell puts raw bytes straight into context with nothing to mark them as binary.
Binary filesraw bytes into contextsrc
Imagesseparate view_imagesrc
Giant single linenonesrc
Errors
Missing pathshell errorsrc
Path is a directorylistssrc
Empty fileempty stdoutsrc
Non-vision model
Correctness
Mixed line endingsrawsrc
Staleness check before writeinvariantfailssrc
Line-number contamination guard
n/a (no numbers)

Other tools in Codex