SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsbq-skill

bq-skill

Office & productivity

View repository

Updated 10-16 08:54

About this skill

BQ Skill

Upgrade from "being able to answer a certain BQ" to "owning a reusable career story bank." Core loop:

挖掘 (Mine) → 结构化 (Structure) → 打标 (Map) → 存库 (Save) → 复用 (Reuse)

First principle: check the bank first, then get to work. Whenever any BQ comes in, first check story-bank/_index.md whether there is a matching story; if it can be reused, reuse/fine-tune it; if not, trigger a new round of mining. This is "the more it is used, the better it understands the user."

Second principle: ask only one question at a time. Mining is a conversation, not a questionnaire. Ask one question → wait for the answer → follow up based on the answer. Never throw out a string of questions at once.

Third principle: do not make up stories for the user. All material must come from the user's real experiences. You can guide, follow up, and help them articulate vague things clearly, but never fabricate Task / Action / Result. Always verify quantified numbers with the user.

Routing: when the user comes in, first determine intent

What the user says Which process to follow
"Help me prepare for an interview" / "I want to build a story bank" / giving a piece of experience Mine new stories → prompts/story-mining.md
Pasting a specific BQ ("Tell me about a time…") Answer one question(check the bank first, process below)
"Is this story of mine told well?" / pasting an existing answer Polish an existing story → prompts/structuring.md
"Mock interview" / "Give me a few questions to test me" Mock interview(v1 lightweight version, below)
Given JD + resume / "Help me prepare BQs for this role" / "What will this company ask, and how should I answer?" JD-driven BQ selection + preparation → prompts/jd-driven-prep.md
"Look at my story bank" / "What stories do I have?" Read story-bank/_index.md report

If you cannot tell, ask one question: "Do you want tomine new stories to build a bank, orprepare for a specific question?"

Mine new stories

Fully execute prompts/story-mining.md the four-layer probing engine in:

  1. Icebreaker layer — specifically for "I do not have any highlights." Use counterfactual questioning + four-quadrant time-anchor scanning to first fish out 3–5 candidate events.
  2. Deep-dive layer — for the selected events, complete STAR one by one, focusing on drawing out the most commonly missing T (what you specifically did, not the team) and R (quantified results)。
  3. Tagging layer — after mining, map competency tags (frameworks/competency-tags.md) + determine which company dimensions they can address (frameworks/company-profiles.md）。
  4. Save into the story bank — write it as a story file according to story-bank/_story-template.md and update _index.md。

Focus one session on mining mining through 1 complete storyis enough; depth matters more than quantity. After mining, ask the user whether they want to continue to the next one.

Answer a specific BQ

  1. Analyze the question: what competency is this question testing? (Refer to frameworks/competency-tags.md to look it up)
  2. Check the story bank: read story-bank/_index.md, and find stories whose tags / competencies match.
    • Match → pull out the story, and reorganize the opening and landing from the angle of this question (the same story can cover multiple questions; see the framework in frameworks/star-car.md）。
    • No match → switch to "Dig for a new story" — dig one up on the spot, then answer.
  3. Produce the answer: default to bilingual Chinese-English — the English is a version you can say directly in an interview, withChinese key pointsfor review.
  4. Conveniently save it into the story bank: if it was newly dug up on the spot, save it into the bank.

Polish existing stories

Run prompts/structuring.md: diagnose the structural problems in the user's existing answer (common ones: Situation too long, can't see what "I" did, no quantified Result, unclear competency tags), and give a rewrite. After rewriting, it can be saved to the bank.

Mock interview

  1. Ask for the target company/role, and load frameworks/company-profiles.md 's corresponding style.
  2. Ask 1 question based on the dimensions that company commonly tests,one question at a time。
  3. After the user answers, give feedback: structure (whether STAR is complete), whether the competency signals are clear, whether quantification is in place, and fit with that company's dimensions.
  4. Prompt the user to save well-answered stories to the bank.

JD-driven BQ question selection + preparation

Fora specific role, do targeted BQ preparation. Fully run prompts/jd-driven-prep.md, a five-step process:

① 上传 JD → ② 上传简历 → ③ 生成 Top 20 选题（对着 JD）→ ④ 基于你的经历给 STAR 准备模板 → ⑤ 输出 HTML 报告
  1. Get the JD — prioritize reusing ../job-description-skill/jd-bank/ 's already-decoded Must Have + Hidden Signals; if there are none, have the user paste the JD and run through the decoding first.
  2. Get the full resume text — tag each experience with competency labels to prepare for matching. If the resume is too thin → switch to "Dig for new stories" and dig a few stories first.
  3. Cross-analysis → Top 20 selected questions — four categories (Behavior general / company-values customized / role-professional / Level-stage customized). Each question must include "what it tests + why this company would ask it(pointing to a certain Must Have / Hidden Signal in the JD) + which story to pair it with".
  4. Prepare templates question by question — by default for the Top 5 must-practice, write complete STAR templates: fill in S/T with resume facts,leave placeholders for A/Rto guide the user to fill in real actions and numbers,never fabricate. Mark that one draft can be used for multiple purposes.
  5. Gap list — truthfully list the questions the JD requires but the resume has no material for, and route them to "Dig for new stories" to dig on the spot.
  6. HTML report — default output to ~/Desktop/Claude skills/bq-prep-<company>-<role>-<YYYYMM>.html(for visual specs see assets/bq-prep-report.md）。
  7. Write back — write the JD↔story mapping back into the job-description-skill's jd-bank file "Predicted Questions", forming a complete job-search chain.

This is the downstream deepening of job-description-skill's interview prediction process: that side gives "what will be asked", and this side gives "how to answer using your real experience + templates」。

Story Bank

  • Location: under this skill's directory story-bank/, one per story .md。
  • metadata relies on frontmatter (tags / competencies / company_fit / metrics / framework / status).
  • _index.md is the reverse lookup table:competency tag → story file, and is the entry point reused by "answering questions".
  • Every time a story is added or modified,be sure to update in sync _index.md。

Reference files (read on demand, don't load them all at once)

  • prompts/story-mining.md — Four-layer probing engine (the core of "digging up new stories")
  • prompts/structuring.md — Structured diagnosis and rewriting ("polishing existing stories")
  • prompts/jd-driven-prep.md — JD × resume → Top 20 topics + preparation template ("JD-driven preparation", linked to job-description-skill)
  • assets/bq-prep-report.md — Visual specification for the HTML report of "JD-driven preparation"
  • frameworks/star-car.md — When to use STAR / CAR / SOAR, and reusing the same story from multiple angles
  • frameworks/competency-tags.md — Competency tag dictionary + BQ question type reverse lookup
  • frameworks/company-profiles.md — Evaluation styles such as Amazon LP / Meta / Anthropic / OpenAI
  • story-bank/_story-template.md — Story file template
  • story-bank/_index.md — Story index