Module pr_review 

Pure-function module for parsing structural PR review comments and evaluating whether a PR meets the auto-merge criteria.

This module intentionally has zero dependency on orchestrator runtime state and performs no I/O. It parses the output of the structural-pr-review Claude Code skill and applies a deterministic policy.

See cto-executive-system/plans/adf-rate-of-change-design.md §Step 1 (Gitea issue terraphim/adf-fleet#29).

Expected review comment layout

The skill emits HTML <h3> section headings (not markdown ###) so that Gitea/GitHub render them consistently. The relevant anchors for this parser are:

• <h3>Summary</h3> — summary paragraphs and optionally acceptance criteria checklist items (- [x] / - [ ]).
• <h3>Confidence Score: N/5</h3> — integer N in 1..=5.
• <h3>Inline Findings</h3> — finding blocks starting with **P0 , **P1 , or **P2 .
• Footer <sub>Last reviewed commit: <short hash> | Reviews (N)</sub> — the | Reviews (N) suffix is optional for first-round reviews.

Structs

AutoMergeCriteria

Thresholds that must be satisfied for a PR to be auto-merged.

PrMetadata

Minimum PR metadata required by evaluate.

ReviewVerdict

Parsed review verdict extracted from a single review comment body.

Enums

AutoMergeDecision

Auto-merge decision produced by evaluate.

VerdictParseError

Errors raised when a review comment body does not conform to the structural-pr-review template.

Functions

author_is_agent

Return true when login belongs to an automation account authorised to open auto-merge-eligible PRs.

evaluate

Evaluate whether a PR clears every auto-merge gate.

parse_verdict

Parse a review comment body emitted by the structural-pr-review skill.