Prompt agents to use edit skill

[samsja]

Note

Low Risk
Prompt-only change gated on installed skills; no runtime or auth behavior changes.

Overview
When the edit skill is installed, the agent system prompt now includes EDIT_SKILL_PROMPT, telling the model to prefer await edit(path=..., old_str=..., new_str=...) for unique exact string replacements and to fall back to normal Python file I/O otherwise. The text spells out supported kwargs and explicitly forbids wrong parameter names and line-based edit styles.

tests/test_prompt.py extends the prompt test helper with an installed_skills override and adds coverage that the edit block appears only when edit is in the installed skills list.

^{Reviewed by Cursor Bugbot for commit 90ba4c9. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 90ba4c9. Configure here.}

[cursor]

Edit prompt without IPython tool

Medium Severity

EDIT_SKILL_PROMPT is appended whenever edit appears in installed_skills, but it instructs agents to call await edit(...), which only works inside the IPython kernel. If RLM_TOOLS omits ipython (e.g. bash,edit), that guidance is injected even though skills are not pre-imported and await cannot run.

 

^{Reviewed by Cursor Bugbot for commit 90ba4c9. Configure here.}