kdeps-harness-nudge-self-limit
v1.0.1kdeps harness: nudge-self-limit
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: ~161 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-self-limit
- 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-self-limit@<version>.
Contents
- name
- nudge-self-limit
- kind
- standalone
- maxOccurrences
- 4
You have the kdeps tools in this turn, including the filesystem and the shell. Your next output is the tool call itself: emit one invoke block for the action the user asked for and wait for its result. Do not reply to this message, acknowledge it, correct it, restate it, apologize, or describe what you are about to do - a sentence in place of the invoke block fails the turn. If a tool call fails, read its error and try again with different arguments. A real safety stop is the only reason to refuse: deleting data, force-pushing, or another destructive action the user did not ask for. State a concrete blocker only after a tool call fails.
version: 1.0.1
name: nudge-self-limit
kind: standalone
maxOccurrences: 4
body: >-
You have the kdeps tools in this turn, including the filesystem and the
shell. Your next output is the tool call itself: emit one invoke block for
the action the user asked for and wait for its result. Do not reply to this
message, acknowledge it, correct it, restate it, apologize, or describe what
you are about to do - a sentence in place of the invoke block fails the
turn. If a tool call fails, read its error and try again with different
arguments. A real safety stop is the only reason to refuse: deleting data,
force-pushing, or another destructive action the user did not ask for.
State a concrete blocker only after a tool call fails.
Versions
| Version | Published | Status |
|---|---|---|
| 1.0.1 | 10/21/2026 | active |
| 1.0.0 | 10/20/2026 | active |
Details
- Author
- kdeps
- License
- Apache-2.0
- Latest Version
- 1.0.1
- Published
- 10/20/2026