| name | cc-optimizer:apply |
| description | Preview or apply per-project config optimizations: model choice, subagent models, cache tuning (dry-run default, use --apply to write). |
| argument-hint | [--apply] [--project slug] [--min-confidence low|medium|high] |
| allowed-tools | Bash, Read, Edit |
/cc-optimizer:apply
Convert findings into real settings.json edits. Safe by default (dry-run); review changes before applying.
Usage
/cc-optimizer:apply # Dry-run: preview all changes
/cc-optimizer:apply --project metropolys # Dry-run: preview one project
/cc-optimizer:apply --min-confidence low --apply # Write all projects (including low-confidence)
/cc-optimizer:apply --project dev-squad --apply # Write one project (with .bak backup)
What it can auto-apply (safe, low-risk)
model: sonnet / model: haiku — switch default model
CLAUDE_CODE_SUBAGENT_MODEL=haiku — cheaper subagents
ENABLE_TOOL_SEARCH=false — defer MCP tool-search
- Cache-prefix stabilizers (
ATTRIBUTION_HEADER=0)
What goes to MANUAL (you decide)
- Force 5m ephemeral cache (
FORCE_PROMPT_CACHING_5M)
- Disable MCP servers per-project
- Other destructive/complex changes
Safety
- Dry-run by default (shows plan, makes no changes)
- Per-file
.bak backups before any write
- Deep-merge: never clobbers other keys
- On-disk verify: confirms intended mutations present
Next step
If changes look good: /cc-optimizer:apply --project slug --apply to write one project.