brd-reconcile

Turn the customer's returned review into frozen decisions, and leave nothing in the tree still asserting a position it overturned: $ARGUMENTS

Core references. A citation of the form workflows-core:<name> names a shared reference in the workflows-core plugin. Load it with Skill(skill: "workflows-core:reference", args: "<name>") — never by path: ${CLAUDE_PLUGIN_ROOT} resolves to this plugin, which does not carry it.

/brd-reconcile is the BRD-to-PRD route's last command (PM phase) — it takes the file that came back from the package /brd-package sent, freezes what the customer actually decided, and then goes looking for everything the delivery organisation still believes that their answer has made false. Its whole discipline is one rule: a customer decision enters the register only when the customer answered and an operator confirmed the answer (D14, ${CLAUDE_PLUGIN_ROOT}/references/decision-register-format.md §1). This command exists to make that happen, not to restate it.

Usage: /brd-reconcile <BRD-KEY> @<review-file> [--sent <path>…]

<BRD-KEY> still resolves through either of the two levels resolve-address searches (workflows-core:addressing §3) — a BRD that owns its source document, or one of its slices — because a root must be resolved before Phase 0 step 5a can refuse it by name. Only a slice is reconciled: a root BRD is refused, and reconciling happens at the slice and nowhere else — a slice holds its own register, its own [C] question set and its own ledger, and it is reconciled from those and no others. Two things follow from inheritance and are named where they arise: its requirement defect resolutions land in its parent's brd/brd-defect-log.md (${CLAUDE_PLUGIN_ROOT}/references/brd-format.md §2.1), and the stale cross-reference sweep's root is the parent's folder, so a sibling slice still asserting a superseded position is reached.

Standing rule, binding on every phase below. Nothing this command reads out of a returned review becomes a [CD#n] on its own. Schema mode parses; free-text mode infers; both hand the result to an operator before a single register record is written, and the operator sees the customer's own words beside every row they are asked to confirm. The section How no inferred decision becomes a [CD#n] below states the ordering that makes this structural rather than aspirational.

This command takes no --no-docs, and it does no documentation grounding at all. That is a decision, not an omission. /brd-intake and /prd-ground already ground this BRD against the shipped product documentation when $DOCS_PATH resolves (D22, workflows-core:docs-grounding), and /brd-interview and /brd-package deliberately do none for the same reason this command does none: it works on decisions already taken — here, on decisions taken by the customer. This command goes one step further than either. A documentation page is a claim about behaviour written by the delivery organisation, and the whole content of this run is what the customer said; consulting a page here could only produce a sentence that contradicts the one party whose authority this command is recording, and then quietly lose the argument to it. So there is no flag to turn off, no resolve-docs-grounding call, and no docs grounding: line in this command's report. The sentence is written here because leaving it unwritten is exactly how the gap it forecloses gets shipped.

No repository is opened, at any point. Every finding this run reads was pinned, written and independently re-derived by /prd-ground, so there is no baseline gate here, no dirty-tree stop, and no $REPOS_PATH requirement. That has one consequence this command states rather than works around: a customer challenge in the review's section 5 or 6, and a will-change finding whose named prerequisite decision this run has just frozen, are both named as needing a grounding pass and are never re-adjudicated here. A finding is not evidence until independently re-derived by a different agent (workflows-core:grounding-format §8), and this command re-derives nothing — a finding it wrote would be an unverified opinion wearing a [CG#n].