Forward Impact Team

Summit Internals

Overview

Summit treats a team as a system, not a collection of individuals. It aggregates individual skill matrices (derived via libskill) into team-level capability views, detects structural risks, and supports what-if staffing scenarios.

Summit is fully local and deterministic -- no external dependencies, no LLM calls.

Architecture

Summit consumes two inputs:

  1. Team roster -- from Map's organization_people table, providing each person's discipline, level, and track
  2. Individual skill matrices -- derived via libskill's deriveSkillMatrix() for each team member

From these inputs, Summit computes team-level aggregations that answer staffing and capability questions for leadership. 

Map (roster) -> libskill (individual derivation) -> Summit (team aggregation)

Skill Matrix Aggregation

For each skill in the framework, Summit computes the team's collective proficiency from individual derivations:

  • Coverage count -- how many team members have each skill at each proficiency level
  • Maximum proficiency -- the highest individual proficiency on the team
  • Distribution -- the spread of proficiency levels across team members
  • Capability coverage -- which capabilities have adequate staffing and which are understaffed

This produces a team skill matrix that mirrors the structure of individual matrices but represents the team's collective capability.

Structural Risk Detection

Summit identifies three categories of structural risk:

Single Points of Failure

A skill where only one engineer is at practitioner level or above. If that person leaves or is unavailable, the team loses its only source of deep expertise in that area.

Critical Gaps

A skill where no engineer is at working level or above. The team lacks the competence to handle this skill area independently -- all work requires external support or guidance.

Concentration Risks

A skill where the majority of engineers are at the same proficiency level. This indicates insufficient knowledge distribution -- the team has breadth but lacks the depth/breadth gradient needed for mentoring and knowledge transfer.

Scenario Modeling

Summit supports what-if simulations that add or remove hypothetical team members and recompute coverage and risks:

  • Add a member -- specify discipline, level, and track for a hypothetical hire and see how team coverage and risks change
  • Remove a member -- simulate a departure and identify which risks emerge
  • Change a role -- model a promotion or track change and see the effect on team capability

Scenarios do not persist -- they are computed on demand and discarded. This enables leadership to explore staffing decisions without affecting real data.

Design Principles

Teams as Systems

Summit measures the team, not individuals. Individual skill matrices are inputs to team aggregation, but Summit never presents individual performance. Leadership sees capability coverage, not individual scores.

Plan Forward, Not Backward

Summit answers "what capability do we need?" not "who is underperforming?" Risk detection identifies structural gaps in the team system, not deficiencies in individual team members.

No External Dependencies

Summit runs entirely locally with no network calls, no LLM inference, and no external services. All computation is deterministic -- the same inputs always produce the same outputs.

Capability, Not Performance

Summit measures what the team can do (capability coverage), not how well they do it (performance). Performance measurement is outside its scope.

Privacy Through Aggregation

Individual skill matrices are aggregated into team views. Summit's output is always at the team level, preserving individual privacy while giving leadership the system-level view they need.

Dependency Chain

libskill -> Summit

Summit depends on libskill for individual skill matrix derivation. It uses the same deriveSkillMatrix() and deriveBehaviourProfile() functions that Pathway uses for job definitions, ensuring consistency between individual and team views.

Map provides the team roster via organization_people, but Summit consumes this through standard data loading rather than activity-layer queries.