菜单
Fit sigmoidal (S-curve) learning curves and forecast future values with the PRA package. Use when a user has historical time/completion (or unit/cost) data showing slow start, rapid acceleration, and plateau, and wants a fitted model plus predictions with confidence intervals.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Compute and update the probability of a risk event from root causes, and turn risk probabilities into project cost distributions, using the PRA package. Use when a user has a risk register with root-cause probabilities and wants a prior risk probability, a Bayesian update after observing causes, or a cost-impact distribution.
Quantify structural coupling between project tasks through shared resources and risks using the PRA package's Design Structure Matrices. Use when a user provides a resource–task matrix (and optionally a risk–resource matrix) and wants to see which task pairs are coupled and therefore tend to fail together.
Measure project cost/schedule performance and forecast final outcomes with Earned Value Management using the PRA package. Use when a user provides a budget (BAC), a planned schedule curve, and actual costs/progress and wants SPI, CPI, variances, or an Estimate at Completion (EAC).
Run Monte Carlo (or fast Second Moment) schedule/cost risk simulations with the PRA package. Use when a user provides task-level uncertainty distributions and needs the range of possible project outcomes — mean, standard deviation, and P5/P50/P95 percentiles — optionally accounting for task correlation.
Route a project risk analysis request to the right PRA tool. Use this as the entry point whenever a user asks about project cost/schedule uncertainty, earned value, risk probabilities, learning curves, or task dependencies and you are unsure which PRA capability applies. Points to the method-specific skills and the underlying MCP tools served by pra_mcp_server().
Size a project contingency reserve and rank the tasks that drive total uncertainty, using the PRA package. Use after (or alongside) a Monte Carlo simulation when a user asks "how much reserve do I need?" or "which tasks should I focus mitigation on?"
| name | pra-learning-curves |
| description | Fit sigmoidal (S-curve) learning curves and forecast future values with the PRA package. Use when a user has historical time/completion (or unit/cost) data showing slow start, rapid acceleration, and plateau, and wants a fitted model plus predictions with confidence intervals. |
The user has historical x (time or unit number) and y (completion, cost,
etc.) data and wants to model the learning curve and project it forward.
fit_and_predict_sigmoidal_tool(x_json, y_json, model_type, predict_x_json, conf_level) fits the model with nonlinear least squares
(Levenberg–Marquardt) and returns fitted coefficients plus predictions.
x_json: JSON array of x values, e.g. [1,2,3,4,5,6,7,8,9,10].y_json: JSON array of y values, e.g. [5,15,40,60,70,75,80,85,90,95].model_type: "pearl", "gompertz", or "logistic".predict_x_json: x values to forecast at; omit to predict at the original x.conf_level: e.g. 0.95 for 95% prediction intervals; omit for none.Fit all three and compare residuals to select the best fit; use confidence intervals to gauge prediction uncertainty (bands widen as extrapolation extends beyond the observed range).