Skip to content
Cadence

/c-check

Source of truth: skills/c-check/SKILL.md. For the why behind the diagnostics, read the Diagnostics deep dive.

Purpose

Section titled “Purpose”

Reviews a design or plan for substance (does it make sense, does it hang together, are there gaps, contradictions, or silent assumptions) and reports back in plain English without touching the doc.

When to use it

Section titled “When to use it”

Use /c-check before approving a design, before running /c-plan or /c-execute, or after a long brainstorm: any time you want a sanity check on whether a doc is ready for the next stage. It is an off-pipeline, read-only diagnostic. Use /c-find-bugs instead when you want concrete code defects, not a “is this good?” judgment.

Invocation forms

Section titled “Invocation forms”
  • /c-check <design-folder-path>: review a full design.
  • /c-check <plan-folder-path>: review a full plan.
  • /c-check <single-doc-path>: review one child doc in isolation.
  • /c-check --format <path>: also run Cadence format-compliance checks (for hand-edited or imported artifacts).
  • /c-check --any <path>: accept paths outside Cadence’s expected folder structure.

Inputs & preconditions

Section titled “Inputs & preconditions”

An existing design or plan doc. It refuses if the path doesn’t exist, or if the path is in a Cadence-managed location but lacks the expected structure (no overview, missing frontmatter), unless --any is set.

What it produces

Section titled “What it produces”

A citation-first report in chat: a plain-English readiness summary, findings grouped Critical / Important / Minor with a fix direction each, a mandatory “What’s working” section, and one specific recommended next action. After printing, it can optionally walk you through applying findings. It never modifies the reviewed doc.

Gates & guarantees

Section titled “Gates & guarantees”
  • Each substance lens (accuracy, internal consistency, completeness, hidden assumptions, scope discipline, internal logic, open-questions) runs as its own parallel sub-agent; plans additionally get codebase-citation and DAG-soundness checks.
  • Severity is applied conservatively: Critical is reserved for things that block the next stage outright.
  • It never reviews code quality beyond verifying a plan’s cited paths/symbols exist, never modifies the doc, never flips status, and never writes a result file (the report stays in chat).
Section titled “Related commands”