メニュー
Run cnbl scan on a target repo to classify every file by kind (domain logic, port, adapter, script, spec, config, glue, discard). Outputs JSONL. Use when the user wants to "scan a repo", "see what's in a project", or "classify files".
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Debug or inspect classifier output for specific files or paths. Shows which rule fired and why. Use when the user asks "why was this classified as X", "what kind is this file", "debug classification", or wants to understand classifier behavior.
Absorb a foreign repo into the Rust ecosystem. Full pipeline: scan, classify, plan routing, generate hexagonal stubs, and execute. Use when the user says "cannibalize", "absorb", "strip", or points at a non-Rust repo to harvest. Orchestrates the scan -> plan -> gen -> eat pipeline with approval gates between stages.
Run cnbl eat to execute a migration plan -- copies generated stubs into destination repos and archives originals to the vault. Use when the user wants to "execute the plan", "absorb the files", "do the migration", or "eat it".
Run cnbl gen to create hexagonal Rust stubs from a migration plan. Generates domain types, port traits, and adapter skeletons. Use when the user wants to "generate stubs", "scaffold components", or "create Rust skeletons".
Run cnbl plan to route classified items to ecosystem repos. Takes scan JSONL as input, matches items against repos.json, and outputs routing decisions. Use when the user wants to "plan migration", "route items", or "decide where files go".
| name | scan |
| description | Run cnbl scan on a target repo to classify every file by kind (domain logic, port, adapter, script, spec, config, glue, discard). Outputs JSONL. Use when the user wants to "scan a repo", "see what's in a project", or "classify files". |
| model | haiku |
| effort | low |
| allowed-tools | ["Bash","Read"] |
Run the scanner on a target repo and display the classification report.
cnbl scan <repo-path> --output /tmp/cnbl-scan.jsonl --report
JSONL stream where each line is a HarvestItem:
{
"rel_path": "src/model.py",
"lang": "python",
"kind": "domain_logic",
"size_bytes": 1234,
"notes": null
}
| Kind | Meaning |
|---|---|
| domain_logic | Types, structs, classes worth porting |
| port | Interfaces, traits, protocols |
| adapter | Infrastructure implementations |
| entrypoint | main files, cmd/ directories |
| test_harness | Test files (discarded during planning) |
| script | Shell/Nu scripts |
| spec | Markdown docs |
| config | TOML/YAML/JSON/BAML config files |
| glue | Re-export files (lib.rs, init.py, mod.rs) |
| discard | Unrecognized or fixture files |
With --report, a summary table prints to stderr showing counts per
kind and the languages found in each category. Use this to get a
quick overview before running plan.