菜单
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?"
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
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 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 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?"
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.
| name | boundary-segmentation |
| description | 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?" |
| allowed-tools | Read |
Sylph's defining engine. Every PostToolUse(Edit|Write|MultiEdit) event
flows into an online agglomerative clusterer. When the next event's
distance to the active cluster exceeds threshold θ (default 0.55), a
task boundary fires — and that triggers the downstream plugins to branch,
commit, and eventually open a PR.
Three components, weighted:
d(event, cluster) = α · (1 − jaccard(files)) α = 0.4 (file overlap)
+ β · (1 − cosine(tokens)) β = 0.4 (content overlap)
+ γ · tanh(idle_gap / τ) γ = 0.2, τ = 300s
Weights and threshold live in shared/constants.sh (SYLPH_BOUNDARY_*).
sylph-learning (W5) tunes them per-developer over time — the defaults
are what a fresh install starts with.
plugins/boundary-segmenter/state/boundary-clusters.json contains:
threshold, uncertainty_band, cfg (alpha/beta/gamma/tau)active: the currently-open cluster (opened_at, events[])closed_clusters: the last 20 finalized clustersEach event in a cluster records files, vector (L2-normalized token
weights), tool, timestamp. The vector shows what tokens the
algorithm matched on — useful for debugging why two edits did or didn't
cluster together.
If Crow is installed and publishes crow.change.classified events on
the mcp-event-bus, Sylph will substitute Crow V1's semantic diff
embedding in place of the stdlib token vector. The API is the same; the
accuracy improves.
A developer asking:
SYLPH_BOUNDARY_* in shared/constants.sh
and sylph-learning/state/learnings.json for per-dev overrides.