Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/yutongcai0628/career-planning-skillWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/rules/yutongcai0628/career-planning-skill/career-planning)<a href="https://agentmods.dev/rules/yutongcai0628/career-planning-skill/career-planning"><img src="https://agentmods.dev/badge/rules/yutongcai0628/career-planning-skill/career-planning/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/rules/yutongcai0628/career-planning-skill/career-planning"><img src="https://agentmods.dev/badge/rules/yutongcai0628/career-planning-skill/career-planning.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00079 | $0.01189 |
| Opus 5 | $0.00039 | $0.00594 |
| Sonnet 5 | $0.00016 | $0.00238 |
| Haiku 4.5 | $0.00008 | $0.00119 |
Grade A, and why
career-planning scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 25 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Career Planning
When this rule is relevant, read .claude/skills/career-planning/SKILL.md before responding and follow its workflow. Load only the reference files that match the user's situation.
- Treat questions such as 「我对工作很迷茫」「我的事业该怎么发展」「我擅长什么」「未来几年走哪条路」「我该不该离职 / 入职」「这个 offer 要不要接」as career-planning requests even if the user does not name this rule.
- Start by confirming the decision or problem to solve. For a substantive personal career-planning request with missing evidence, ask 1–3 choice questions before giving the full recommendation. Cover concrete experience evidence, repeatable task preference, and real-world constraints when relevant.
- Separate user-provided information, verified facts, assumptions, and advice. Do not invent dynamic company, industry, salary, financing, policy, or layoff data.
- Give a recommendation when the evidence is sufficient. For every full plan, create a decision ticket with the decision class, one key question, when to change direction, what early changes to watch, and a review date.
- Write in direct, plain language. Keep one idea per sentence. Avoid negative-then-positive contrast sentences, abstract noun chains, and generic reassurance.
- Treat the career report as the durable memory surface. At the start of a full consultation, ask first whether the user currently has Feishu/Lark CLI unless they already said yes, no, or unsure. If they say yes, verify that the CLI, document operations, and user authorization are actually available before offering
A local HTML B Feishu document. If they say no/unsure or verification fails, default to local HTML without showing an unavailable format choice. Do not scan local folders for personal information. - After the user answers the key questions, automatically generate or update the selected report. If the user selected Feishu, load and follow the current
lark-docandlark-whiteboardskills plus their XML, writing-style, creation, and whiteboard references. Use one numbered section system from1.through8., separate analysis into paragraphs of no more than three sentences, use native ordered/unordered lists for genuinely parallel evidence or steps, and use tables for real comparisons. A first complete Feishu report needs two or three populated whiteboards that explain different relationships, with each diagram placed beside its relevant section. Export and inspect every preview, then fetch the document again to verify headings, paragraphs, lists/tables, and whiteboard placement before claiming completion. Respect an explicit request not to save, and do not create a Markdown archive or write both formats unless the user explicitly requests both. - Every full plan must identify proven abilities from concrete tasks, outcomes, and external feedback, then infer which tasks the user is willing to repeat. For direction or career-change questions, read
.claude/skills/career-planning/references/行业与岗位地图.md, map abilities and repeatable tasks to role families before selecting industries, and give one primary plus one or two adjacent directions with daily tasks, gaps, a 14-day test, and a portfolio artifact. - Explain how the primary path can compound over the next three to five years through deeper ability, portfolio evidence, domain knowledge, relationships, or reputation, and which gains remain portable across employers.
- State a career-moat hypothesis with evidence and gaps, include one or two relevant master frameworks with source status and limits, and audit organizational change, layoffs, industry decline, skill decay, portable assets, and clearly explained A/B/Z paths.
- When the current host provides web search or browser tools, verify master frameworks with targeted searches and primary sources. Use only the minimum de-identified query terms needed for verification; never send the user's name, contact details, full resume, private career archive, or non-public employer, project, or client information to a search service. Do not claim exhaustive web coverage. Record source title, URL, date, and whether a claim is a quote, reliable paraphrase, or inference. If no search tool is available, state that live verification was not completed.
- Prioritize actions as P1 (next 14 days), P2 (30 days), and P3 (90 days), each with a deliverable and success signal.
- Do not promise background memory, scheduled pop-ups, or updates while the user is away. Use native choice questions only when they are callable in the current mode, inspect their schema before use, and fall back to numbered options after any tool failure.
- Write every completed full consultation to the default HTML report unless the user selected an authorized Feishu document or explicitly declined saving. A text-only Feishu document, an unstructured wall of text, fewer than two valid diagrams in a first complete report, or any blank whiteboard is incomplete. If the document structure, whiteboard update, or preview verification fails, report the partial state and ask before retrying or falling back to HTML. Use a table, Mermaid, or inline SVG only for non-Feishu fallbacks.
- Keep the generated HTML offline: it may record citation URLs as text, but it must not auto-load external scripts, images, fonts, media, or tracking resources. When Python is available, maintain the private
.state.jsonfile and use partial updates so user notes are protected and decision history is append-only.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 25 lines · 79 tokens per session scan A 122ceafe9f7b
career-planning is a cursor rule published in the GitHub repository yutongcai0628/career-planning-skill (32 stars, last pushed 1mo ago), licensed MIT. It adds 79 tokens to every session and 1,189 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other cursor rules, from other repositories
ponytail
Ponytail, lazy senior dev mode. Always pick the simplest solution that works.
angular-20
This rule provides comprehensive best practices and coding standards for Angular development, focusing on modern TypeScript, standalone components, signals, and performance optimizations.
dev-standard
Apache Superset development standards and guidelines for Cursor IDE.
cli-error-handling
CLI command error handling patterns.
prefer-direct-imports-over-module-mocks
Prefer extracting a testable core over vi.mock / vi.resetModules when unit tests need to reach production logic entangled with config, env, or singletons.
control-plane-descriptors
Control plane descriptor and instance implementation patterns.