Tool Result
Hasil Tool
What the harness sends back after running a tool: file contents, command output, an error.
A tool result is the payload that goes back into the context after a tool runs. It is usually the most important text in the whole context, because it is the only part describing the actual state of the world.
Its size matters more than people expect. Running `ls` on a large directory and pasting everything back can consume a large slice of the context window before the agent has done any work. Truncating results, or making the tool return a summary instead of a dump, is often the single biggest context win available.
Results are also untrusted input. A file that contains text telling the agent to ignore its instructions is data, not instruction, and a harness that pastes it in without distinction has handed the agent a way to be redirected.
Related terms
Tool Call
The model's request naming a tool and its arguments. Structured output, not a side effect.
Tool
A function the model is allowed to call, like reading a file or running a command.
Context Engineering
Deciding what goes into the model's context, and what stays out.
Prompt Injection
Text that tries to make the model ignore its instructions and follow the attacker's instead.
Agent Loop
The repeated cycle of decide, act, observe. Where autonomy actually comes from.
Want this working inside your team?
It starts with one conversation about your workflow.