kdeps-harness-nudge-sandbox-hallucination
v1.0.0kdeps harness: nudge-sandbox-hallucination
set harness / shelf harness
PurposeThe text kdeps sends the model in agent mode: the sections of the system prompt, and the prompts behind compaction, goals, judges, prompt refinement and nudges.
LLM context: ~118 tokens when loaded (prompt or tool definition)
Where it runs
Part of the harness set, used in workflow and agent mode. The text kdeps sends the model in agent mode: the sections of the system prompt, and the prompts behind compaction, goals, judges, prompt refinement and nudges.
- you type a prompt
- the agent loop builds the request
- the harness sections become the system promptharness/nudge-sandbox-hallucination
- the model answers
- the loop runs tools or replies
Loaded by: kdeps (agent loop), /settings, kdeps run, kdeps konfig export, kdeps knowledge build, kdeps konfig import
In the code: kdeps (agent loop)agent.Newagent.harnessText
Install
A kdeps asset: installed with the first-run profile that includes its set, or the first time a command needs it, and updated without a new release.
Pin a version with kdeps install harness/nudge-sandbox-hallucination@<version>.
Contents
- name
- nudge-sandbox-hallucination
- kind
- standalone
- maxOccurrences
- 2
You made no tool call, and phrases like "NO CONTENT AVAILABLE", "expired", "/mnt/data", or "cannot access the filesystem" come from a code-interpreter sandbox that is not this environment. kdeps has no sandbox and never returns those messages -- nothing ran. The real working directory is a live filesystem with the files from the task present now. Make an actual kdeps tool call (bash_exec, read_file, ...) through the tool interface and wait for the runtime's result.
version: 1.0.0
name: nudge-sandbox-hallucination
kind: standalone
maxOccurrences: 2
body: >-
You made no tool call, and phrases like "NO CONTENT AVAILABLE", "expired",
"/mnt/data", or "cannot access the filesystem" come from a code-interpreter
sandbox that is not this environment. kdeps has no sandbox and never
returns those messages -- nothing ran. The real working directory is a
live filesystem with the files from the task present now. Make an actual
kdeps tool call (bash_exec, read_file, ...) through the tool interface and
wait for the runtime's result.
Versions
| Version | Published | Status |
|---|---|---|
| 1.0.0 | 10/11/2026 | active |
Details
- Author
- kdeps
- License
- Apache-2.0
- Latest Version
- 1.0.0
- Published
- 10/11/2026