Updated system_prompt.md to align with the current runtime toolset and discovery workflow.
Context
The prompt needed a tooling refresh so instructions map directly to current behavior and available tools.
Changes
Updated tool inventory to the current set.
Replaced multi-step search guidance with a discover-first workflow.
Enforced discover-only routing for repository search, lookup, and exploration.
Added explicit prohibition on using bash for repository searching.
Added an end-of-prompt final reminder block that repeats discover-for-search and bash-not-for-search guidance.
Updated examples to demonstrate discover-first behavior and follow-up file reads/edits.
Updated penalties to enforce use of the declared tool list.
Clarified path usage for file tools and reinforced parallel batching rules.
Behavioral Impact
The model receives unambiguous guidance aligned to current tool availability.
Repository searching is now explicitly routed through discover.
End-of-prompt repetition reinforces that bash is for execution tasks, not code search.
Reduced risk of invalid tool calls from ambiguous search instructions.
1---2name: 2500-2026-02-20-system-prompt-tooling-refresh-3e8dd37e3description: Summary4---56## Summary7Updated `system_prompt.md` to align with the current runtime toolset and discovery workflow.89## Context10The prompt needed a tooling refresh so instructions map directly to current behavior and available tools.1112## Changes13- Updated tool inventory to the current set.14- Replaced multi-step search guidance with a discover-first workflow.15- Enforced discover-only routing for repository search, lookup, and exploration.16- Added explicit prohibition on using bash for repository searching.17- Added an end-of-prompt final reminder block that repeats discover-for-search and bash-not-for-search guidance.18- Updated examples to demonstrate discover-first behavior and follow-up file reads/edits.19- Updated penalties to enforce use of the declared tool list.20- Clarified path usage for file tools and reinforced parallel batching rules.2122## Behavioral Impact23- The model receives unambiguous guidance aligned to current tool availability.24- Repository searching is now explicitly routed through `discover`.25- End-of-prompt repetition reinforces that `bash` is for execution tasks, not code search.26- Reduced risk of invalid tool calls from ambiguous search instructions.
Run npx skillmds@latest add tools-only/2500-2026-02-20-system-prompt-tooling-refresh-3e8dd37e in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.
Summary It is listed under AI & ML on SkillMD.
This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.
This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.
Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.
tools-only (@tools-only) published this skill. Their other Agent Skills are listed on their SkillMD profile.
tools-only