| name | ask |
| description | Ask. Achieve. The canonical user-facing entry to claudemax — describe your goal, the SOTA-2026 pipeline (deepresearch → multispec → parallel /goal → verify) handles the rest. Same engine as /cmax with a friendlier verb. CLI equivalent is `cmax ask "<goal>"`. |
/ask — describe your goal, achieve it
The user-facing entry point. You describe what you want; the pipeline produces it.
CLI equivalent: cmax ask "<your goal>" — same engine as /cmax and cmax run.
What runs (auto, no flags)
- Taste check — read
taste.md + taste.vision if present. Suggest cmax taste init if absent and the repo has signal.
- /deepresearch — web-current research with source ledger persisted to
memory.research_sources.
2-bis. verify-research (default ON) — decomposed per-claim verification: parallel citation-support checks on cached excerpts, network escalation only on doubt, harness-owned timeouts (timeout = unverified, never pass). Contradicted findings stripped before decompose; failed briefs re-researched once.
- multispec decompose — Fable 5 (while included →2026-06-22; Opus after) authors a MultiSpec: N sub-Specs + DAG + rollup completion conditions + per-sub-Spec writeSet.
- /specqa (parallel per sub-Spec) — gate spec quality (every verifyHint mechanically checkable).
- /introspect (parallel per sub-Spec) — confidence/assumption gate. Blocks at confidence < 6.
- Mode selection — Mode A (SDK subagents in one query()) for ≤ 5 sub-Specs / short runs; Mode B (Claude Code Agent Teams with shared task list + worktree isolation) for big swarms or write-set overlap.
- Parallel /goal per DAG leaf — Sonnet by default; Opus on router escalations (security, novelty, complexity ≥ 7).
- Per-sub-Spec /verify (parallel, blind Opus) — re-reads repo, runs verifyHints.
- Rollup /verify — blind Opus checks integration conditions across all sub-Spec outputs.
- Memory record + state snapshot. ntfy.sh push to phone if
NTFY_TOPIC is set.
Bundled dark-patterns hooks block vibes, fake citations, aggregator hallucination, and credential leaks throughout.
Variants (when you need to deviate)
/ask runs /opussonnet semantics by default (Opus plan/judge/verify + Sonnet exec). For other shapes, use the explicit skill:
/opusolo <goal> — all-Opus exec for novel/security/payments/auth (~3× cost, max effectiveness)./cmax <goal> — same as /ask; the brand umbrella./workflow <goal> — v1 muscle-memory alias./opussonnet <goal> — explicit v1 muscle-memory invocation.
CLI overrides on cmax run:
cmax run "<goal>" --variant opusolo
cmax run "<goal>" --mode teams
cmax run "<goal>" --no-research
cmax overnight SPEC.md --budget-credits 50
When NOT to invoke /ask
- Trivial one-line changes — just edit.
- A single SPEC that fits in one /goal loop — invoke
/goal directly on the existing SPEC.md.
- Pure research with no implementation — invoke
/deepresearch directly.
