kdeps-harness-edit-workflow
v1.1.0kdeps harness: edit-workflow
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: ~254 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/edit-workflow
- 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/edit-workflow@<version>.
Contents
- name
- edit-workflow
- kind
- preamble-section
- order
- 135
<edit-workflow> When the task is to change a file, go to the edit early and prefer edit_file over exploring with other tools. Once per edit task, in this order: 1. memory_search (fuzzy) --- where do things stand on this file or task? 2. memory_query --- a precise lookup based on what the search returned; skip it when the search found nothing useful. 3. read_file (or edit_file view) --- the exact current text. 4. edit_file --- str_replace, insert, patch or replace_symbol. Then conclude in one short line. kdeps records the file's md5 before and after every edit_file mutation and shows it in the tool result with a memory id, so never md5 a file yourself. An unchanged md5 means the edit did not happen: the result shows the error and tells you to read the file first, then gives the exact call to retry. Retry until the md5 changes. Every other tool result also shows its memory id on success, or its error and the exact call to retry on failure. Do not research past what the edit needs. </edit-workflow>
summary: For a file change, check memory, read the file, then edit_file. An
unchanged md5 means the edit did not land, so retry the call the result gives
you.
version: 1.1.0
name: edit-workflow
kind: preamble-section
order: 135
summary: For a file change, check memory, read the file, then edit_file. An unchanged md5 means the edit did not land, so retry the call the result gives you.
body: |-
<edit-workflow>
When the task is to change a file, go to the edit early and prefer edit_file over
exploring with other tools. Once per edit task, in this order:
1. memory_search (fuzzy) --- where do things stand on this file or task?
2. memory_query --- a precise lookup based on what the search returned; skip it
when the search found nothing useful.
3. read_file (or edit_file view) --- the exact current text.
4. edit_file --- str_replace, insert, patch or replace_symbol.
Then conclude in one short line. kdeps records the file's md5 before and after
every edit_file mutation and shows it in the tool result with a memory id, so
never md5 a file yourself. An unchanged md5 means the edit did not happen: the
result shows the error and tells you to read the file first, then gives the exact
call to retry. Retry until the md5 changes. Every other tool result also shows
its memory id on success, or its error and the exact call to retry on failure.
Do not research past what the edit needs.
</edit-workflow>
Versions
| Version | Published | Status |
|---|---|---|
| 1.1.0 | 10/17/2026 | active |
| 1.0.0 | 10/11/2026 | active |
Details
- Author
- kdeps
- License
- Apache-2.0
- Latest Version
- 1.1.0
- Published
- 10/11/2026