Skill flags
Every --flag accepted by the four d2c skills. This page reads the ## Arguments section from each upstream SKILL.md at request time, so it stays in sync without a doc rewrite every time upstream adds a flag. Defaults are what the skill uses when the flag is omitted.
d2c-init
| Flag | Default | Description |
|---|---|---|
--force | — | Skip incremental update detection and force a full scan (Steps 0-8). Useful if the design-tokens.json is out of sync and incremental detection fails. |
d2c-build
| Flag | Default | Description |
|---|---|---|
--threshold <number> | 95 | Pixel-diff match percentage required to pass the visual verification gate. Must be between 50 and 100. Example: `--threshold 90` accepts 90% match. |
--max-rounds <number> | 4 | Maximum number of visual verification rounds before stopping. Must be between 1 and 10. Example: `--max-rounds 6` allows up to 6 fix iterations. |
d2c-audit
| Flag | Default | Description |
|---|---|---|
--fix | — | After reporting, auto-fix hardcoded values that have an exact token match (labeled `definite` in Audit 1). Only fixes exact matches — `undocumented` values are reported but not touched. Show a summary of fixes applied after the report. When combined with `--dry-run`, shows a preview of proposed fixes instead of applying them (see below). |
--dry-run | — | Run the full audit and generate the report, but do NOT apply any changes. Behavior depends on whether `--fix` is also provided: - **`--dry-run` alone** (without `--fix`): Runs the audit as read-only and generates the report. This is the same as the default behavior but makes the read-only intent explicit. - **`--dry-run` with `--fix`**: Runs the full audit, identifies all fixable issues, but instead of auto-fixing, presents each proposed fix as a diff-style preview: ``` --- src/components/Header.tsx (current) +++ src/components/Header.tsx (proposed) @@ line 42 @@ - color: #3B82F6 + color: theme.colors.primary ``` After showing all proposed fixes, ask the user how to proceed: - **"Apply all"** — Apply every proposed fix. - **"Review individually"** — Step through each fix one by one. For each, the user can approve or skip. - **"Skip all"** — Discard all proposed fixes and end the audit. Group fixes by file for easier review. Show the total count: "Found X fixable issues across Y files." |
d2c-guard
No flags — this skill is hook-triggered, not invocable.
See the token schema for the contract these flags operate on, or the per-skill page on the sidebar for the full SKILL.md.