kdeps-harness-handshake-evidence-retry
v1.0.0kdeps harness: handshake-evidence-retry
set harness / shelf guardrails
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: ~147 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/handshake-evidence-retry
- 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/handshake-evidence-retry@<version>.
Contents
- name
- handshake-evidence-retry
- kind
- standalone
That didn't come back as a real tool result -- no problem. To be clear about what's different here: a normal assistant session often has its own built-in way to run code or fetch a result, triggered outside the text you write. kdeps does not give you that channel. It is a plain parser reading your message for a matched <invoke> block, so the block itself has to be the literal text you send -- open tag, parameters, close tag, nothing else around it and no markdown code fence. Send exactly this now: <invoke name="bash_exec"> <parameter name="command">pwd</parameter> </invoke>
version: 1.0.0
name: handshake-evidence-retry
kind: standalone
body: >-
That didn't come back as a real tool result -- no problem. To be clear
about what's different here: a normal assistant session often has its own
built-in way to run code or fetch a result, triggered outside the text you
write. kdeps does not give you that channel. It is a plain parser reading
your message for a matched <invoke> block, so the block itself has to be
the literal text you send -- open tag, parameters, close tag, nothing else
around it and no markdown code fence. Send exactly this now:
<invoke name="bash_exec">
<parameter name="command">pwd</parameter>
</invoke>
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