Decision Analysis

Expose assumptions, failure paths, success conditions, and option-changing evidence before a consequential design choice is recorded.

This is a reference skill used by discuss and adr-writer. It analyzes; it does not create artifacts, mutate lifecycle state, write implementation, create Work Items, or perform VCS operations.

When Analysis Helps

Use deeper analysis when plausible options have non-obvious trade-offs, the choice is expensive to reverse, consequences cross important boundaries, or a wrong assumption could create normative, compatibility, data, security, or operational risk.

Use a compact comparison for reversible, well-understood choices. Skip this skill when an existing normative RFC already determines the answer or when the task is a routine implementation choice with no durable architectural consequence.

Hard Stops

• Do not treat invented scenarios, unsupported probabilities, or model confidence as evidence.
• Do not recommend behavior that conflicts with a normative RFC.
• Do not turn an ADR into the source of new product obligations.
• Do not perform an implementation experiment before the governing lifecycle permits implementation work.
• Stop and clarify when an unknown could materially reverse the option ranking or violate an authority boundary. State lesser assumptions explicitly.

Analysis Policy

Frame The Choice

Identify the decision, viable options, governing constraints, success criteria, relevant horizon, and available evidence. Include the status quo only when it is a real option. Separate facts from assumptions and note which evidence would change the ranking.

Premortem

For each viable option, imagine a concrete failure at the relevant horizon. Identify the smallest set of material internal and external causes, their early signals, likely impact, and feasible prevention or response. Include specification drift, hidden coupling, migration or data loss, verification blind spots, operational burden, and governance cost only when relevant.

Backcast

Imagine the same option succeeding. Work backward to the prerequisites, controllable actions, external dependencies, and measurable signals that made success possible. This tests whether the option has a credible path, not merely an attractive outcome.

Compare And Recommend

Rank options against governing constraints, downside severity, reversibility, evidence quality, cost of learning, and success prerequisites. Use quantitative probability or impact values only when supported by data; otherwise use clear relative judgments with confidence and rationale.

Return Go, No-Go, or Conditional Go, the preferred option, why it wins, conditions that must hold, evidence that would trigger a pivot, and the smallest lifecycle-valid step that reduces the most important uncertainty. Prefer analysis, interface sketches, prototypes outside governed behavior, or isolated experiments that do not cross an RFC phase boundary.

Scale the depth to decision risk. Do not generate a large matrix when a short comparison exposes the decisive trade-off.

Authority Mapping

Analysis output is input to governed artifacts, not a new source of authority:

Output	Destination
Observable obligations, compatibility rules, validation behavior	RFC
Context, evaluated alternatives, chosen direction, consequences	ADR
Approved delivery scope, acceptance evidence, dependencies, mitigation follow-up	Work Item
Stable repeatable executable checks	Verification Guard
Assumptions, scenario detail, temporary evidence	Discussion record or final response

Summarize only durable rationale in the ADR. Do not paste the full analysis, execution gates, or task checklist into its decision field. Use adr-writer to shape the selected content and preserve structured alternative ownership. A pivot that could change the selected decision or a requirement returns to discuss and the owning ADR or RFC before execution resumes.

Completion Evidence

A useful analysis:

• states the decision, viable options, constraints, evidence, and assumptions;
• gives each serious option a credible failure and success path;
• identifies signals and mitigations that can change action;
• makes a decisive, risk-calibrated recommendation without false precision;
• proposes a lifecycle-valid uncertainty-reduction step; and
• routes obligations, rationale, execution, and verification to their authoritative destinations.