KelvinYou

@kelvinyou source repo

13 skills
0 followers
11 hours ago last updated

▌
Meta Coach · kelvinyou
Monthly audit of the advice quality from weekly-review and coach-planner — not an audit of the user's behavior. Trigger when the user says "audit the agent's advice", "meta audit", "meta-coach", "how good has the agent's advice been", "is coach's advice reliable", or "why do I keep missing P0". Requires ≥ 4 weekly reports + ≥ 4 timetables to have enough material to analyze.

0
installs
▌
Daily Report · kelvinyou
Turn the user's brain dump (free-form natural-language notes) into a structured daily review log that conforms to Personal-OS conventions. Use when the user describes what they did today, how much they spent, their sleep, etc.

0
installs
▌
Decision Log · kelvinyou
Capture non-trivial decisions into the Personal-OS decision log, supporting a quick one-line brain dump or an interactive walkthrough. Trigger when the user says "log that I decided...", "help me log a decision", "log this decision", "I decided to...", "decision-log", or any scenario where the user clearly expresses that they made an important trade-off/choice. Does not give decision advice (that's coach-planner's job) — only records decisions.

0
installs
▌
Coach Planner · kelvinyou bundle
Personal-OS's sole scheduling agent: owns all timetable generation (daily/weekly/next-week), plus real-time action advice and decision support. Plans schedules based on recent log data, circuit breaker status, current body/energy state, and the P0/P1/P2 objectives produced by weekly-review. Trigger when the user asks "how should today be arranged", "schedule next week", "should I skip the morning run", "does this week's plan need adjusting", "help me schedule today/tomorrow/this week/next week", "I'm behind on deep work", "should I skip training today", "plan my day", "plan next week", or any question about action advice and scheduling. Trigger even for a casual "what should I do right now". Do not confuse with weekly-review (the weekly review report) — weekly-review only produces the diagnostic report + objectives; coach-planner owns all schedule execution.

0
installs
▌
Weekly Review · kelvinyou
Generate a comprehensive weekly review report for Personal-OS: aggregate daily logs, score across 4 dimensions (Output/Health/Mental/Habits), enforce circuit breakers, compare week-over-week trends, and output next-week P0/P1/P2 objectives (but NOT timetables — timetable generation is coach-planner's job). Use this skill whenever the user mentions weekly review, weekly report, week summary, "how was my week", wants to review their week, or says "make report" / "make weekly". Also trigger when the user asks about weekly scores, trends, sleep debt trajectory, or wants to generate a W## report. Do NOT trigger for "schedule next week" or schedule requests — those go to coach-planner.

0
installs
▌
Identity Audit · kelvinyou
Quarterly identity audit: infer "the top 3 things that mattered most last quarter" from behavioral data, and compare against the claims in data/user_profile.md. Trigger when the user says "quarterly audit", "identity audit", "what did my quarter actually look like", "quarterly review", or "behavior vs. claims". Requires ≥ 12 weeks of log data.

0
installs
▌
Learning Agent · kelvinyou bundle
AI-era personal skill radar + hiring market scan: generates a structured skill learning checklist, priority ranking, and skill-demand/salary analysis based on real MY/SG hiring data for the user. Triggers when the user asks 'what should I learn recently', 'is there any new tech worth watching', 'help me update my learning plan', 'what skills are most valuable right now', wants to understand tech trends / upskilling / self-improvement / career development / learning paths, or asks 'what skills are employers hiring for', 'what's the MY/SG salary for role XX', 'help me scan the job market', 'which skills raise pay the most', 'what are AI roles / SWE roles hiring for right now', or wants to understand real hiring demand and emerging role requirements. Should also trigger even if the user just casually mentions a new tech term and wants to know if it's worth learning, or mentions wanting to see the market landscape for some role category.

0
installs
▌
Wealth Manager · kelvinyou bundle
Analyze stocks (Bursa Malaysia & US markets), identify buy-the-dip opportunities, manage investment portfolio, and summarize tracked assets across all supported vehicles (stocks, MMFs, FDs, digital banks). Use this skill whenever the user asks about stocks, portfolio, investments, savings allocation, interest rates comparison, net worth, asset allocation, a financial/investment plan, tax-efficient investing (PRS, ETF withholding tax), or mentions buying/selling shares — even if they don't explicitly say "wealth" or "portfolio". Also trigger when the user provides a new trade, updates savings placement, asks "where should I put my money", or "is my plan any good".

0
installs
▌
Decision Review · kelvinyou
Review decision log entries that are due: compare expected vs. actual outcome, assess calibration error, extract lessons. Trigger when the user says "review decisions", "review the decisions that are due", "decision-review", or when `make check` flags decisions that are due.

0
installs
▌
Contract Guardian · kelvinyou bundle
Review semantic contracts in Personal-OS whenever a change touches schemas, templates, scripts/lib, thresholds, wealth reports, Python/TypeScript boundaries, public/private data paths, agent skills, or architecture and plan documents. Use this skill whenever the user asks whether a cross-layer change is complete, whether a schema or field rename is safe, whether docs and code agree, whether data may be public, or whether a report/skill contract needs tests or migration—even if they do not say "contract" or "schema". Build an owner → consumer → migration/test impact map, classify drift with the project's Status vocabulary, and block unsafe semantic changes. This skill reviews contracts; repo-orchestrator owns multi-repo release order and git-commit owns explicit staging and commits.

0
installs
▌
Profile Optimizer · kelvinyou bundle
Optimizes the user's LinkedIn / Jobstreet / portfolio content based on real MY/SG hiring data: analyzes skill gaps, rewrites experience bullets, and gives section-ordering and trimming recommendations. Triggers when the user asks 'help me optimize LinkedIn', 'how should I change my profile', 'how should my resume be laid out', 'portfolio ordering', 'check my profile before I apply to this role', 'how should my jobstreet profile read', 'is my headline any good', wants recruiters/HR to notice them more easily, or wants to rewrite their own content against a benchmark profile. Should also trigger even if the user just pastes a chunk of their own profile and wants feedback. Do not confuse with learning-agent — learning-agent fetches JD data and identifies what to learn; profile-optimizer consumes that data to rewrite your profile text and layout.

0
installs
▌
Repo Orchestrator · kelvinyou bundle
Coordinate multi-repository development in Personal-OS, especially changes that span the main repo, private data, ai-stock-analysis, portfolio-website, market data, or agent skills. Use this skill whenever the user asks to sync or bump submodules, prepare a release or integration check, review whether a change is safe to commit, finish work across repositories, inspect dirty submodule state, or verify the correct order of tests and parent-pointer updates—even when they do not explicitly say "orchestrate" or "release". This skill prepares and validates the integration; explicit commit or push requests hand off to git-commit after the checks pass.

0
installs
▌
Quant Backtest Review · kelvinyou bundle
Adversarially review new or changed backtest, scoring, and signal-generation code in repos/ai-stock-analysis before its output is trusted for a real trade. Use this skill whenever the user asks to review a backtest change, add a new strategy or feature to the backtest/scorer/portfolio modules, interpret a backtest report's numbers, decide whether a hit rate or Sharpe is "real", or asks "does this signal actually work" / "can I trust this number" — even if they don't say "backtest" or "statistics" explicitly. Also trigger before any wealth-manager decision that cites a backtest result the user hasn't had checked yet. This skill checks for lookahead bias, overfitting, unrealistic transaction costs, and statistical significance problems (deflated Sharpe, multiple testing, small samples) — it does not do general code review or multi-repo integration checks.

0
installs

13 published skills