review-rep-call-trends

bởi openai

Phân tích các cuộc gọi gần đây của nhân viên bán hàng hoặc nhân viên tiếp xúc khách hàng để phát hiện các mô hình cải thiện, suy giảm và ổn định với bằng chứng khách quan và hướng dẫn đào tạo thực tế…

npx skills add https://github.com/openai/role-specific-plugins --skill review-rep-call-trends

Review Rep Call Trends

Context-Gathering Intake

Whenever this skill asks for context, strongly prefer using the answers-ask-user-input skill and the ask_user_input tool over other tools such as request_user_input; otherwise ask directly in the conversation.

Turn a representative time-based call sample into objective coaching guidance. This skill compares older and newer call evidence to identify genuine improvement, regression, and stable patterns. It owns the readout only; it does not create trackers, send feedback, post messages, or save artifacts.

Common Skill Instructions

MANDATORY: If not already in context, read and adhere closely to plugins/sales/skills/index/SKILL.md## Cross-Skill Best Practices.

Key Dependency Categories

These categories are particularly important for this workflow; use other sources only when they materially improve sampling or interpretation.

  • [Blocking] ~~Meeting Transcripts for time-sliced call search, summaries, transcript moments, speaker context, dates, companies, and clickable call links. It blocks the default live-source trend path; explicit older/newer call evidence already in context satisfies the need.
  • ~~Knowledge & Files for user-provided or exported transcripts, call summaries, manager-provided call sets, and prior coaching context

Call summaries or transcripts are required for trend claims. Manager notes, CRM outcomes, and generic impressions may provide context, but cannot substitute for older-versus-newer call evidence. If the sample is small or skewed, return a limited trend readout rather than overgeneralizing.

Reference Loading

SKILL.md owns the normal time-sliced trend workflow and output format. Load references only when their extra detail matters:

Workflow Guidance

1. Resolve the rep and comparison basis

  • If the target rep or comparison basis is not clear, ask the user via ask_user_input(). For a bare invocation such as “review call trends,” do not assume a rep or time window; offer: My calls over the default trailing 90 days, A named rep over the default trailing 90 days, and Supplied older/newer call sets or a custom window.
  • If the user names only “my call trends” without a comparison basis, offer the same explicit choices before broad retrieval. Do not infer a representative rep, comparison window, or trend-review mode from a fresh request.
  • Require a target rep or clearly supplied target call set and a comparison basis: time window, older/newer sets, transcript set, call summaries, or explicit coaching dimension.
  • After the user selects “my calls,” use the current user when available. For an ambiguous manager-style request, ask for the rep's exact display name or email before broad retrieval.
  • If an attendee-filtered search returns zero results, try small grounded variations: first/last name only, then an obvious nickname when context supports it. If those still return zero, ask for the rep's exact display name in the selected ~~Meeting Transcripts source.
  • After the user selects the default Trend Review path, use the trailing 90 days split into three non-overlapping slices: 0-30 days, 31-60 days, and 61-90 days. Convert every relative window to concrete absolute dates before searching.
  • Preserve user-specified motion, product focus, account, coaching dimension, or audience. If a supplied date phrase could materially change the sample, ask one brief clarification.
  • If the target remains ambiguous after the user selects a lane, make at most three narrow source reads, offer up to five concrete candidates, and confirm any inferred anchor before deeper analysis.

2. Collect summaries across time

  • Start with ~~Meeting Transcripts and search each time slice separately. Use the rep's exact email or display name, a limit of about 25 per slice, and score_threshold=0 when supported—or the lowest practical threshold—so the sample is not biased toward only “strong” calls.
  • When the user names a focus such as discovery, demo, renewal, objections, or technical troubleshooting, use a short matching query. Otherwise use an empty or neutral query to avoid biasing the sample.
  • Aim for 30-60 deduplicated calls when available, with a mix of call types and companies visible in summaries. If a slice is sparse, expand it once by about 30 days and state the adjustment.
  • If there are hundreds of calls, prioritize the most recent calls plus a smaller older baseline while preserving visible call-type and company variety.
  • Build the trend view from summaries first: date, company, title, topics, call link, and the retrieval handle kept only for fetching.
  • Do not claim improvement or regression from a single recent call or from outcomes alone.

3. Compare supported behavior categories

  • Compare the same categories across older and newer slices whenever possible.
  • Use the core categories: opening/agenda, discovery depth, qualification/prioritization, positioning/narrative, objection handling, next steps and mutual action plan, technical accuracy, clarity/concision, executive presence, and listening/turn-taking.
  • When the calls are technical or solution-oriented, use the rubric's additional categories for problem framing, diagnosis/troubleshooting, solution design/tradeoffs, and technical-plus-business stakeholder management.
  • Prefer three to six categories with the clearest evidence. Mark mixed evidence as Inconsistent rather than forcing a trend.
  • Call something Improved or Regressed only when the time comparison supports a directional shift. Use Stable for repeated patterns without clear movement.

4. Fetch a small validation set

  • Fetch full context only where summaries are ambiguous, a trend needs validation, or a concrete coaching moment would help.
  • Normally inspect one to two strong recent calls, one to two recent calls with friction, and one to two older baseline calls.
  • Capture two to six brief moments: what happened, why it mattered, category, readable title/date/company context, and live link.
  • Keep quotes short and use the full content to validate—not replace—the summary-level sample.

5. Turn trends into action

  • Tie each improvement, regression, and stable pattern to evidence from the compared slices.
  • Recommend exactly three coaching actions for the next two weeks. Each action needs a skill to practice, an if/then play, a five-to-ten-minute drill, and an observable check for the next calls.
  • Select three to eight calls to re-listen to, favoring the clearest recent and baseline examples that explain the coaching priorities.

Next Step Options

After the first output, offer the most relevant follow-up from the options below. Offer one clear transition, not a menu. Suggest ONLY these unless you are very confident another option is more useful:

  • Drill into one improved, regressed, or inconsistent behavior category.
  • Turn the top three actions into a two-week practice plan or coaching note for review.
  • Compare a bounded peer-exemplar set for one coaching priority.
  • Select and summarize the most useful calls to re-listen to.
  • Set up a future trend refresh for the same rep, scope, and comparison basis.

Next steps to avoid:

  • Making performance judgments from thin evidence or sending feedback automatically.

Modes

  • Trend Review — default trailing-window analysis.
  • Focused Trend Review — use for a named motion, category, account set, or supplied older/newer calls.
  • Limited Trend Readout — use when one slice, transcript depth, or sample balance is insufficient for confident movement claims.

Output Format

Return these sections in order.

# Rep Call Trends: [Target Rep]

## Coverage

- **Target rep:** [Name]
- **Window analyzed:** [Absolute start] -> [Absolute end]
- **Time slices:** [Older / middle / recent absolute ranges]
- **Calls reviewed:** [N summaries], [M full fetches]
- **Sample shape / limits:** [Mix, sparse slice, or confidence note]

## What Improved

- **[Behavior change]** — [Evidence from older vs newer readable call context + inline links] — [Why it matters]

## What Regressed / Risk Signals

- **[Behavior change]** — [Evidence from older vs newer readable call context + inline links] — [Why it matters]

## What Stayed Consistent

- **[Stable or inconsistent pattern]** — [Evidence + inline links]

## Top 3 Coaching Actions: Next 2 Weeks

1. **Practice:** [Skill]
   - **If / then:** [In-the-moment play]
   - **Quick drill:** [5-10 minute drill]
   - **Look for next:** [Observable signal]

## Calls to Re-Listen

- [Readable date, company, title + inline link] — [Why this call matters]

---

{Follow the instructions and output format/conditions in [Limitations and Improvements](../index/SKILL.md#limitations-and-improvements)}

{Follow the instructions and output format/conditions in [Next Steps](../index/SKILL.md#4-next-steps)}

Example Prompts

  • Review a rep's call trends.
  • Review Jamie's call trends over the last month.
  • Compare this rep's recent discovery calls to their earlier calls and flag changes.
  • Find improvement and regression patterns across this rep's demo calls.

Rules

  • Do not invent calls, behavior, outcomes, metrics, quotes, dates, companies, or links.
  • Every improvement or regression claim must compare older and newer evidence and cite at least one readable call link; do not turn anecdote into trend.
  • Prefer explicit structure, questions, summaries, objections, and next steps over speculative interpretation. Do not infer intent or personality.
  • Cite with readable title/date/company context and compact inline numbered Markdown links, such as 2026-03-24, ExampleCorp, Pilot & Pricing [1](https://example.com/call/123). Prefer live transcript or call URLs; never expose raw connector ids.
  • When one statement is supported by multiple calls, append multiple numbered links inline.
  • Keep excerpts short, normally no more than 25 quoted words from one source.
  • If a useful call has no stable link, use readable context and say that no direct link was available.

Failure Handling

If the target or comparison basis cannot be resolved, state the smallest missing input and offer concrete candidates when available. If the sample is sparse or uneven, produce the limited readout, name which movement claims are unsupported, and suggest the smallest additional call set or window that would improve confidence.

Thêm skills từ openai

user-context
openai
Tải hoặc quản lý các tùy chọn định tuyến nguồn bền vững, logic giới thiệu, tiến trình thiết lập và sổ đăng ký lớp ngữ nghĩa của plugin Phân tích Dữ liệu.
official
notion-research-documentation
openai
Nghiên cứu nội dung Notion và tổng hợp thành các bản tóm tắt có cấu trúc, báo cáo hoặc so sánh kèm trích dẫn. Tìm kiếm và truy xuất các trang Notion bằng truy vấn mục tiêu, sau đó sắp xếp kết quả theo chủ đề với trích dẫn nguồn trong văn bản và phần tài liệu tham khảo. Chọn từ bốn định dạng đầu ra (tóm tắt nhanh, tổng hợp nghiên cứu, so sánh, báo cáo toàn diện) dựa trên phạm vi và mục tiêu của người dùng. Tạo và cập nhật các trang Notion bằng mẫu có sẵn; liên kết trực tiếp nguồn và theo dõi thay đổi khi
official
rcsb-pdb-skill
openai
Gửi yêu cầu RCSB PDB nhỏ gọn để lấy siêu dữ liệu cốt lõi, truy vấn API Tìm kiếm và tải xuống FASTA. Sử dụng khi người dùng muốn tóm tắt RCSB ngắn gọn; lưu JSON thô hoặc…
official
pdf
openai
Đọc, tạo và xác thực PDF với kết xuất trực quan và tạo theo chương trình. Kết xuất các trang PDF sang PNG để kiểm tra trực quan bố cục, khoảng cách và kiểu chữ trước khi bàn giao bằng Poppler (pdftoppm). Tạo PDF theo chương trình với reportlab để định dạng đáng tin cậy; trích xuất văn bản và siêu dữ liệu bằng pdfplumber hoặc pypdf. Thực thi các tiêu chuẩn chất lượng: không có văn bản bị cắt, phần tử chồng lấn, bảng bị hỏng hoặc hiện vật kết xuất; chỉ sử dụng dấu gạch nối ASCII, trích dẫn dễ đọc cho con người. Sử dụng...
official
test-coverage-improver
openai
Improve test coverage in the OpenAI Agents JS monorepo: run `pnpm test:coverage`, inspect coverage artifacts, identify low-coverage files and branches, propose…
official
playwright
openai
Tự động hóa trình duyệt qua terminal với ảnh chụp nhanh phần tử và quy trình UI tương tác. Hoạt động thông qua script wrapper playwright-cli (yêu cầu npx); hỗ trợ chế độ headless và headed để gỡ lỗi trực quan. Quy trình cốt lõi: mở trang, chụp nhanh để tham chiếu phần tử ổn định, tương tác bằng refs, chụp lại sau khi điều hướng hoặc thay đổi DOM. Bao gồm điền biểu mẫu, nhấp chuột, gõ văn bản, quản lý nhiều tab, chụp ảnh màn hình/PDF và ghi lại trace để gỡ lỗi luồng. Tham chiếu phần tử (ví dụ: e3, e15)...
official
ukb-topmed-phewas-skill
openai
Lấy các bản tóm tắt PheWAS UKB-TOPMed nhỏ gọn cho các biến thể đơn lẻ bằng cách chấp nhận đầu vào rsID, GRCh37 hoặc GRCh38 và phân giải thành truy vấn GRCh38 cần thiết. Sử dụng khi một…
official
code-review-context
openai
Ngữ cảnh hiển thị của mô hình
official