docs(roadmap): add #454 — claw list/run/exec/ask fall through to prompt dispatch; suggest_similar_subcommand KNOWN_SUBCOMMANDS too narrow

[Yeachan-Heo]

ROADMAP pinpoint #454 — typo-suggestion candidate list is too narrow

Dogfooded for the 2026-05-24 06:30 Clawhip pinpoint nudge (message 1507994083769974825).

The asymmetry

$ claw ls   # caught as typo because "skills".contains("ls")
[error-kind: unknown] error: unknown subcommand: ls.
  Did you mean     skills

$ claw list  # falls through to LLM prompt dispatch
[error-kind: missing_credentials] error: missing Anthropic credentials; ...

The missing_credentials envelope proves "list" was sent to Claude as a user prompt.

Code

rust/crates/rusty-claude-cli/src/main.rs:1324-1364:

fn suggest_similar_subcommand(input: &str) -> Option<Vec<String>> {
    const KNOWN_SUBCOMMANDS: &[&str] = &[
        "help", "version", "status", "sandbox", "doctor", "state",
        "dump-manifests", "bootstrap-plan", "agents", "mcp", "skills",
        "system-prompt", "acp", "init", "export", "prompt",
    ];
    // Levenshtein ≤ 2, prefix match ≥ 4, or substring match
    ...
}

list has Levenshtein ≥ 3 from every entry, prefix < 4 with every entry, and no substring overlap → returns None, typo arm at :983-994 is skipped, falls through to CliAction::Prompt.

ls only works by coincidence ("skills".contains("ls")). Any refactor of KNOWN_SUBCOMMANDS could silently regress that.

Other natural CLI-discovery shapes that fall through

Spelling	Today	Should be
claw ls	✅ typed typo (coincidence)	typed typo
claw list	❌ prompt dispatch	typed typo → skills list/mcp list/agents list
claw run hello	❌ prompt dispatch	typed typo → prompt
claw exec hello	❌ prompt dispatch	typed typo → prompt
claw ask what	❌ prompt dispatch	typed typo → prompt / REPL
claw models	❌ prompt dispatch (#452)	CliAction::Models once #452 lands
claw providers	✅ guard	(covered by #453 once extra-args fix lands)
claw profiles	❌ prompt dispatch	typed typo
claw aliases	❌ prompt dispatch	typed typo
claw logs	❌ prompt dispatch	typed typo
claw settings	❌ prompt dispatch	typed typo

Why distinct from #78 / #145 / #452 / #453

• 见证历史 #78 / 合影 #145 filed claw plugins as a single missing CliAction variant.
• avicii #452 (PR docs(roadmap): add #452 — claw models/list/help/--help falls through to prompt dispatch (prompt misdelivery, sibling of #78) #3065) filed claw models* as a single missing CliAction variant.
• 合影 #453 (PR docs(roadmap): add #453 — slash-command guard fires only on bare token; any extra arg falls through to prompt dispatch #3066) filed the slash-command guard being shape-blind to subcommand args.
• This pinpoint is the typo-suggestion candidate list being too narrow to catch natural cross-tool CLI shapes. Operators from gh, kubectl, ls, npm, pip, git, aws, openai api will reach for these spellings.

Required fix shape (full detail in ROADMAP entry)

(a) Extend KNOWN_SUBCOMMANDS to include list, ls, models, providers, profiles, aliases, logs, settings, run, exec, ask, chat, auth, login, logout with explicit did_you_mean mappings. (b) Replace Levenshtein/prefix/substring coincidence with explicit did-you-mean targets. (c) Structured kind:"unknown_subcommand" JSON envelope with prompt_dispatch_blocked:true. (d) Regression coverage for every spelling in the table.

Acceptance check (one-liner)

env -u ANTHROPIC_API_KEY -u ANTHROPIC_AUTH_TOKEN claw list should NOT exit with missing_credentials; it should exit with unknown subcommand: list. Did you mean skills list / mcp list / agents list?.

—
[repo owner's gaebal-gajae (clawdbot) 🦞]