pra-overview

name	pra-overview
description	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().

What PRA offers

The PRA R package exposes its analytical methods as tools over the Model Context Protocol via pra_mcp_server(). Once the server is registered, an agent can call these tools directly. Each analytical area also has a dedicated skill with fuller guidance.

Choosing the right tool

The user has...	Use MCP tool	Skill
Task distributions (normal/triangular/uniform), wants total cost/duration	mcs_tool	pra-monte-carlo
Only task means + variances, wants a fast estimate	smm_tool	pra-monte-carlo
Monte Carlo results, wants a reserve buffer	contingency_tool	pra-sensitivity-contingency
Task distributions, wants the biggest risk drivers	sensitivity_tool	pra-sensitivity-contingency
BAC + schedule + actual costs, wants performance/forecasts	evm_analysis_tool	pra-earned-value
Root-cause probabilities, wants risk probability	risk_prob_tool / risk_post_prob_tool	pra-bayesian-risk
Risk cost impacts, wants a cost distribution	cost_pdf_tool / cost_post_pdf_tool	pra-bayesian-risk
Historical time/completion data, wants a forecast curve	fit_and_predict_sigmoidal_tool	pra-learning-curves
A resource–task (and optional risk–resource) matrix	parent_dsm_tool / grandparent_dsm_tool	pra-dsm

Common multi-step workflows

1. Schedule/cost risk: mcs_tool → contingency_tool + sensitivity_tool
2. EVM analysis: evm_analysis_tool (computes PV, EV, AC, SV, CV, SPI, CPI, EAC, ETC, VAC, TCPI in one call)
3. Bayesian prior → posterior: risk_prob_tool → cost_pdf_tool → [observe causes] → risk_post_prob_tool → cost_post_pdf_tool
4. Learning curve: fit_and_predict_sigmoidal_tool

Conventions

• All array/matrix arguments are passed as JSON strings (e.g. task_dists_json, cor_mat_json). Pass "null" (or omit) an optional correlation matrix to treat tasks as independent.
• Never fabricate numerical results — call the tool.
• The probabilistic-network functions (prob_net, prob_net_sim, prob_net_learn, prob_net_update) are available in the R package but are not exposed as MCP tools; use them from an R session.