菜单
Explains how W4 Path-History Reviewer Routing ranks candidate reviewers — recency-weighted blame, path-depth specificity, CODEOWNERS boost, and optional Crow reviewer-availability filtering. Helpful when a developer asks "why did Sylph suggest these reviewers?"
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | reviewer-router |
| description | Explains how W4 Path-History Reviewer Routing ranks candidate reviewers — recency-weighted blame, path-depth specificity, CODEOWNERS boost, and optional Crow reviewer-availability filtering. Helpful when a developer asks "why did Sylph suggest these reviewers?" |
| allowed-tools | Read |
For each candidate (git log author OR CODEOWNERS entry):
score = (sum over changed-paths of: recency_weight(last_commit) × path_depth_weight(path))
× (1.5 if CODEOWNERS matched for any changed path else 1.0)
× availability (0.0..1.0; default 1.0 when unknown)
Where:
recency_weight(ts) = exp(-age_days × ln(2) / 90) — 90-day half-life.path_depth_weight(path) = 1.0 + min(0.5, depth × 0.1) — deeper paths
are worth more than root-level changes, capped at 1.5×.Top-K is capped by SYLPH_REVIEWER_MAX_SUGGEST (default 3) to avoid the
Kubernetes-style "review storm" anti-pattern.
Parsed from (in order): .github/CODEOWNERS, CODEOWNERS,
docs/CODEOWNERS. Sylph uses the GitHub CODEOWNERS glob dialect:
* matches within a path segment** matches across segments/ anchors to root/ matches directory + contentsLast matching entry wins (the spec's "most specific rule wins" is implemented as "last matching line wins" per GitHub's actual behavior).
CODEOWNERS candidates are unioned with blame-graph candidates, not
merged on top. An @-handle from CODEOWNERS and a "Name " from
blame can both be suggested — the PR adapter (W4) does de-dup at the
gh pr create --reviewer level.
If Crow is installed and publishes crow.reviewer.availability.changed
events (availability ∈ [0.0, 1.0] per user), W4 multiplies the score by
availability before ranking.
Without Crow, all candidates are assumed fully available (availability = 1.0). Sylph does not try to parse out-of-office signals from git logs or calendar integrations.
Kubernetes' early test-infra bot auto-assigned reviews to 8+ maintainers per
PR, fragmenting attention. They eventually shipped OWNERS_ALIASES +
randomized-subset selection. Sylph skips to the fix: cap at 3, rotate
across the larger eligible pool on subsequent PRs on the same paths
(rotation logic lives in W5 via state/reviewer-history.jsonl).
gh pr edit <n> --add-reviewer <handle> is fine — W4's suggestion is a
suggestion. The reviewer-history.jsonl tracks overrides so W5 can learn
"this developer always adds @tech-lead regardless of blame" and adjust
future suggestions.
/sylph:pr proceeds with no reviewers; the PR
opens as draft without assignments. Reviewers can be added manually.[]. That's fine for a genuinely new area.Surfaces the decision-gate to the user when a destructive git operation is detected, pausing the session for explicit confirmation before the command runs. Invoked by sylph-gate's PreToolUse hook when a Bash call is classified destructive or protected-destructive.
Explains how W2 Jaccard-Cosine Boundary Segmentation works, what events fire it, and how to interpret the cluster state in plugins/boundary-segmenter/state/boundary-clusters.json. Invoked when a developer asks "why did Sylph commit/branch here?" or "what's W2 doing?"
Explains how Sylph composes PR descriptions — the 4-section template, the fallback ladder when Crow V4 continuity or W2 cluster data is missing, and how to override the default template via PULL_REQUEST_TEMPLATE.md.
Explains the W3 decision tree — how Sylph chooses between GitHub Flow, Trunk-Based, GitFlow, Release Flow, Stacked Diffs, and Unknown. Helpful when a developer asks "why did Sylph pick this branch name?" or wants to override detection via .sylph/workflow-map.yaml.