escalation plugins

by Mohit Aggarwal

Customer Success skills: Customer Health Scorecard, QBR Deck, Escalation Brief, Churn Analysis. Score account health with a weighted RAG framework, build structured QBR decks with value narratives, write crisp escalation briefs for at-risk accounts, and analyse churn by category and segment with prioritised interventions.

by Mohit Aggarwal

Support & customer-ops skills: Support Macro, Help-Center Article, CSAT/NPS Analysis (with an NPS calculator), Support Runbook, Escalation Tree, and KB Audit. For frontline support, CX, and success teams.

by Matt Corbett

Technical-program-management team — agents (technical-program-manager, cross-team-dependency-manager, program-launch-coordinator) for delivering a program of projects across teams: program charter + measurable outcomes, dependency mapping and critical-path/integration risk, a RAID-driven program plan, launch-readiness reviews with written go/no-go criteria, staged rollout, and exec status comms that lead with decisions and asks (not activity). skills, a knowledge bank with Mermaid decision trees (TPM-vs-PM-vs-EM, escalate-or-not, go/no-go), best-practices, templates, commands, and an advisory anti-pattern hook. Distinct from project-management (a single project/plan) and engineering-management (people/teams); execution lands in devops-cicd, observability-sre, and the relevant engineering plugins. Requires ravenclaude-core@>=0.7.0.

by cjthompson

Reduced-toolset sub-agent profiles (read-only, lean-executor, standard-executor, main, full-executor) that lower System-tools token overhead vs. spawning the default agent. Pair with the project-tasks skill, which dispatches scouts, executors, and verifiers to these agents.

by Jordan Richlen

Classify a candidate action as PATCH/MINOR/MAJOR (semver-style blast-radius test) before acting — act silently on PATCH, flag-and-stage MINOR, stop for explicit human sign-off on MAJOR. Use whenever you're mid-task and unsure how much autonomy to take on the next action: which of several implementation paths to pick, whether to overwrite unreviewed state, whether to disable a safety toggle, or any judgment call settings.json's autoMode patterns don't enumerate.

by Oratta

意思決定移譲のフレームワーク「観点の配役」の実装プラグイン。catalog/catalog.md: 3グループ14観点・列定義・横断軸・変更手続き・変更記録を持つ観点カタログの唯一の正本。catalog/injection.md: 14観点の注入タイミング（9分類）・配線先機構・注入文書（.claude/casting/policies/）の置き場所規約を持つ注入マップの正本。skills/casting: 配役表（project.md/local.md/セッション宣言の3層）の作り方・判例台帳の書き方・カタログ変更手続きの手順正本。commands/init.md（/casting:init）: 対象 git repo に `.claude/casting/` 一式を雛形から生成し、gitignore に local.md を冪等追記する。scripts/casting-check.sh: 配役表の5列ちょうどに割れない壊れた表行・閉じられていない HTML コメント・閉じられていない Markdown のコードフェンス・未知の観点語彙・「カタログ外」判例・同一観点の「論点じゃなかった」2件以上・catalog_version 不一致・相談判例（経路「相談の上自走した」）の事後報告5要素欠落の8項目を検出する lint。resolve サブコマンドは有効配役表を合成して出力する（検証を通らない配役表では出力せず exit 1 の fail-closed）。rules/perspective-casting.md（このプラグインとは別ファイル）が毎ターン効く返信前チェック5手順を持つ薄い常時ロード層。　agents/casting-specialist.md（model: sonnet、tools: Read/Grep/Glob/WebFetch — 全文未把握の外部規約を判断前に参照先で読む）・agents/casting-arbiter.md（model: fable）: 論点相談・仲裁プロトコル（#127）の汎用サブエージェント。スペシャリストは指定された policies/<slug>.md を読み人格ブロックを纏って意見を返し、仲裁はフェーズ宣言文と主張リスト（メインセッションの主張1件＋各人格の主張1件ずつ）のみを入力に裁定する。templates/policy.md: 注入文書（人格ブロック＋判断基準＋前提とする外部規約）の雛形。catalog/delegation.md: 委任の定義正本（委任＝許可ツール × 任された観点。ロールは組み合わせ名。`## 委任` 配下に許可ツール表と任された観点表を並べる宣言書式。宣言は要約で正本は permission 設定と配役表 3 層）。commands/policy-interview.md（/casting:policy-interview <観点>）: 観点の判断基準を主と 1 問ずつやり取りして policies/<slug>.md を生成・更新する。templates/delegation.md: /casting:init が生成する委任宣言の雛形。

by Mike Sherrard

PLUGIN 9 — Cohort Claude Support. The calm, always-available support desk inside Cowork. One phrase — "help" — and it takes over: answers every Claude question (Cowork vs Chat, connectors, plans and limits, permission prompts, sessions), diagnoses breakage with real look-only checks down proven decision trees (brain missing, connector expired, skill won't fire, Claude slow), teaches concepts in plain realtor English, audits setup and resumes at the first missing step, tracks the member's cohort week with a catch-up plan, and — when a human is needed — builds the complete ticket ready to paste straight into Mike's support portal. Knows all 9 plugins deeply and routes every fix to the skill that owns it; support itself is READ-ONLY by constitution. Stays current by design: money and limit answers are always fetched live from official pages, and a weekly digest cycle reads Anthropic's changelogs so answers stand on this week's Claude. Requires Plugin 1 (the Brain) for personalization; helps even without it.

by Fars Saadeldin

AI-powered customer support agent trained on the company's own manuals and FAQs — pre-sales questions and support tickets alike: audits the help centre, restructures it so the agent answers without hallucinating, reports coverage gaps, and defines which questions must reach a person.

by Hodios contributors

Hodios incident entries: prompts, personas and workflows.

by Fars Saadeldin

Structured non-clinical check-ins with escalation always on: routes to real resources, never diagnoses, never handles crisis, and produces supervisor summaries and an audit log.

by yukineko

Stuck-loop detector + escalation for Claude Code: a PostToolUse hook that spots repeated identical actions and edit thrash, then injects an escalating nudge to change approach or ask the user for help. Subscription-native (one hook + bundled Rust binary, no API key).