cxgrd prompt

This command works for Pro, Team and Enterprise plans.

Generate a deterministic, architecture-aware prompt for your coding assistant. CXGRD combines your change request with the dependency graph and blast-radius results in ./cg, listing focused files, per-file impact details, and actionable verification steps. Prompt generation runs locally and does not require an LLM API key.

Usage

cxgrd prompt "description of change"

Example output

PS C:\Users\name\project1> cxgrd prompt "renaming get_installation_token to get_token "
🔄 Generating deterministic coding prompt (Pro)...
   Input: renaming get_installation_token to get_token

✓ Generated prompt

Implement: renaming get_installation_token to get_token

Refactor/change focus:
- `src/auth/token.ts` — direct change target identified from the request; architecture layer: service.
- `src/routes/installations.ts` — high risk, direct dependent; Depends on token.ts; review and update if the change requires it.
- `src/cli/commands.ts` — medium risk, transitive dependent, depth 2; Depends on installations.ts; review and update if the change requires it.

Blast-radius summary: medium overall risk; 1 direct target(s), 2 affected dependent(s).

Actionable steps:
1. Inspect the direct targets and their listed dependents; preserve existing public interfaces and dependency direction unless the request requires changing them.
2. Make the requested change in the focused files, updating dependent files only where the blast-radius notes indicate compatibility work is needed.
3. Run the targeted unit and integration tests for the affected files. Include relevant boundary, error-handling, and regression cases, then run the project test suite.
4. Run `cxgrd check --changed` and review any new architecture or dependency issues.

   deterministic engine · plan: pro
   Saved to C:\Users\name\project1\.cg\last-prompt.md