Route code-exploration requests to gitnexus, cx, claude-mem, Read, or Grep. Use when locating a symbol, tracing callers or impact, exploring an unfamiliar module, planning a rename/refactor, checking change scope, or recalling a past decision. Not for plain-text/log/doc search, non-code questions, or a tool already chosen. Output: one primary tool, exact command or MCP call, and a named fallback.
Scanned 9/2/2026
Install to Claude Code
npx -y skills add hmj1026/dhpk --skill dhpk-tool-routing --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Dhpk Tool Routing?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/hmj1026-dhpk-tool-routing-dhpk)More formats (shields.io, HTML) on the badges page.
---
name: dhpk-tool-routing
description: 'Route code-exploration requests to gitnexus, cx, claude-mem, Read, or Grep. Use when locating a symbol, tracing callers or impact, exploring an unfamiliar module, planning a rename/refactor, checking change scope, or recalling a past decision. Not for plain-text/log/doc search, non-code questions, or a tool already chosen. Output: one primary tool, exact command or MCP call, and a named fallback.'
metadata:
dhpk-invocation-class: implicit-eligible
---
# Tool Routing
Route one code-exploration intent to one primary tool and one fallback. The compact card at `${CLAUDE_PLUGIN_ROOT}/rules/tool-routing.md` is authoritative for row order and tie-breakers.
## Route
1. Read `${CLAUDE_PLUGIN_ROOT}/rules/tool-routing.md` sections "Decision tree" and "Tie-breakers" before selecting a tool.
2. Map the request to exactly one intent row. Use its primary tool when available; use its named fallback when the primary is unavailable or returns no useful result.
3. For a planned edit, rename, or refactor, read the card's "gitnexus_impact timing" guidance before touching an existing symbol.
4. Return the route contract below, then stop routing so the caller can run the selected tool.
Completion criterion: the request has one matched intent, one executable primary call, and either a concrete fallback call or an explicit `none`.
## When NOT to Use
- Run the already-selected tool directly.
- Send plain-text, log-line, comment, and documentation searches to `Grep` directly.
- Read one known small file in full with `Read`.
- Answer non-code questions from the relevant configuration or prose source.
## Conditional references
- For an ambiguous request, overlapping intent rows, or a primary tool that returns no result, read `references/decision-tree.md` for edge cases and rationale.
- When dispatching a sub-agent that will inspect code, read `${CLAUDE_PLUGIN_ROOT}/rules/tool-routing.md` "Sub-agent prompts" and paste the relevant block from `${CLAUDE_PLUGIN_ROOT}/docs/subagent-prompt-template.md` into its prompt.
- For impact timing or the append-only exemption, read `${CLAUDE_PLUGIN_ROOT}/rules/tool-routing.md` "gitnexus_impact timing" and the Glossary in `${CLAUDE_PLUGIN_ROOT}/rules/execution-policy.md`.
- Before using a rename or broad search, read `${CLAUDE_PLUGIN_ROOT}/rules/tool-routing.md` "Top anti-patterns".
## Output
```text
Intent: <matched decision-tree row>
Primary: <exact command or MCP call>
Fallback: <exact command or MCP call, or none>
```
## Verification
- [ ] Read the decision tree and tie-breakers before routing.
- [ ] Selected exactly one intent row, primary call, and fallback or `none`.
- [ ] Chose `cx` over `gitnexus` when both are valid, except for GitNexus-only impact, rename, and change-detection operations.
- [ ] For edits to an existing symbol body or signature, satisfied the `gitnexus_impact` timing rule.
- [ ] Began symbol-level source lookup with the selected semantic tool before broad `Read` or `Grep`.
Read `references/decision-tree.md` only for the conditional cases above.
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!