Content Review

Skill id: content-review

Generated from JSON Schema — do not edit by hand. Run bun run cat-harness-tools/scripts/gen-schema-docs.ts.

Input

Input schema for formal content review and approval workflow.

Field	Type	Required	Description
reviewType	"l2-to-l3-gate" | "l3-to-publish-gate" | "change-assessment" | "final-signoff"	yes	Type of review phase gate.
contentRef	string	yes	Git ref or path to the content being reviewed.
previousVersion	string	no	Git ref of the previous approved version (for diff).
validationReport	string	no	Path to the validation report for this content.
stagingLinks	any	no	The staging preview of the work under review, as the request for feedback carried it, so the reviewer opens the changed pages rather than rebuilding links. Rule: cat-harness skills/sdlc/sdlc-core/staging-review.md §”The deep-link rule”.

Raw schema · ✎ Edit

Output

What a review produces: findings, a decision about them, and the audit notes that justify it. A finding is an observation; a decision is an act about a set of findings; neither determines the other. Mirrors schemas/qa-review.ts — see skills/sdlc/sdlc-core/decision-audit.md.

Field	Type	Required	Description
findings	array	yes	One entry per observation. Immutable once raised: to disagree with one, overrule it in the decision.
decision	object	yes
notes	array	yes	Why the decision was made, and why each overruled finding was left.
stagingLinks	any	no	The staging preview of the work this output asks a person to look at: one deep link per changed page, with the same page on main. Required whenever a preview exists; when it does not, say so in missing. Rule: cat-harness skills/sdlc/sdlc-core/staging-review.md §”The deep-link rule”.

findings[]

Field	Type	Required	Description
id	string	yes
subject	string	yes	Block label, skill name or repo-relative path.
criterion	string	no
reviewer	object	yes
severity	"critical" | "major" | "minor"	no	Machine axis — what kind of breakage. Required of a script reviewer.
weight	"blocking" | "suggestion" | "praise"	no	Human axis — what the reviewer asks of the gate. Required of a human reviewer. praise has no image under the machine axis.
detail	string	yes
evidence	array	no
raised_at	string (date-time)	no

findings[].reviewer

Field	Type	Required	Description
kind	"script" | "agent" | "human"	yes
id	string	yes
version	string	no

decision

Field	Type	Required	Description
outcome	"approve" | "request-changes" | "reject"	yes
role	string	yes	Role id from schemas/role-graph.ts — the lane this was decided in.
by	object	yes
considered	array	yes	Finding ids weighed, overruled or not.
overrules	array	no	Findings the decision went against. The finding stays on the record; the note says why.
notes	array	yes	Audit-note ids. At least one.
decided_at	string (date-time)	no

decision.by

Field	Type	Required	Description
kind	"script" | "agent" | "human"	yes
id	string	yes

decision.overrules[]

Field	Type	Required	Description
finding	string	yes
note	string	yes

notes[]

Field	Type	Required	Description
id	string	yes
author	object	yes
rationale	string	yes
cites	array	yes	A rationale with no citation is an assertion.
proposed_by	string	no	The agent that assembled the citations. Never the author — searching the corpus is clerical work.
written_at	string (date-time)	no

notes[].author

Field	Type	Required	Description
kind	"script" | "agent" | "human"	yes
id	string	yes

Raw schema · ✎ Edit