| name | solution-space-convergence |
| description | Apply the Solution-Space Convergence Framework to transform a completed problem-space analysis (typically output from the Problem-Space Exploration skill) into a justified, auditable solution specification. Use this skill whenever the user explicitly requests a solution-space analysis, invokes the framework by name ("solution-space convergence", "SSC framework", "convergence framework"), or asks to "run SSC" on a given problem analysis. Also trigger when the user says "converge on a solution", "generate and select solution candidates", "run a trade study", "build a concept-selection matrix", "design rationale capture", or references specific framework phases (C0–C7) by name. Do NOT trigger on generic architecture discussion, coding tasks, or ad-hoc trade-off conversation unless the user explicitly names this framework. Requires a prior problem-space analysis as input; if none exists, redirect to the problem-space-exploration skill first.
|
Solution-Space Convergence Framework — Skill
First Step: Load the Framework
Before doing anything else, read the full framework specification:
view /path/to/this/skill/references/framework.md
The framework document is the authoritative source for all phase
definitions, notations, formats, and rules. Do not rely on memory or
summaries — load it fresh every time.
Precondition: A Problem-Space Analysis Must Exist
This framework does not generate its own problem analysis. It
consumes one. If the user has not produced a problem-space analysis,
redirect them to the problem-space-exploration skill first and stop.
Acceptable inputs for this framework are:
- A completed PSE analysis document (preferred).
- A problem statement with explicit mandatory properties, constraints,
and at least one narrative execution path, produced by any method
that separates problem from solution with comparable rigor.
If neither is available, do not proceed — any "solution" produced from
an unstable problem is built on sand.
Execution Modes
This skill supports two modes. Ask the user which mode they want
before proceeding, unless they have already indicated a preference.
Mode A — Full Execution
Claude produces a complete solution-space convergence analysis in a
single pass.
When to use: The user provides a PSE analysis (or equivalent) and
wants a finished convergence document. Typical prompt: "Run SSC on this
problem analysis."
Process:
- Read the problem-space analysis (input).
- Verify the input meets the precondition (see "Input Adequacy Check"
in the framework document). If not, stop and report the gap.
- Execute Phases C0 through C7 sequentially. Phase ordering is strict —
see "Phase Ordering Rules" below.
- Apply the Reverse-Contamination Guard (Phase C0) at the end of every
phase. Flag any solution decision that silently redefines the problem
and either reject it or escalate it as a re-analysis event.
- Produce a single Markdown document as output (see "Output Format").
Mode B — Guided Facilitation
Claude walks the user through each phase interactively.
When to use: The user wants to reason through solution candidates
collaboratively, or the problem analysis has ambiguities that require
human judgment to resolve. Typical prompt: "Let's converge on a
solution for this problem."
Process:
- Start by confirming the problem-space analysis is fixed for the
duration of the convergence. Any mid-flight change to the problem
is a re-analysis event, not an edit.
- Walk through each phase in order. For each phase:
- Explain what the phase does and why it matters (briefly).
- Ask targeted questions drawn from the framework's required
dimensions.
- Synthesize the user's answers into the phase output.
- Run the Reverse-Contamination Guard on the output.
- Confirm the phase output before advancing.
- After all phases are complete, compile into the final Markdown
document (see "Output Format").
Phase Ordering Rules
Phase ordering is not optional. The phases have dependency
relationships:
- Phase C0 (Reverse-Contamination Guard) is established first and
applied continuously throughout all subsequent phases.
- Phase C1 (Input Adequacy Check) must pass before any candidate
generation. Generating candidates from an inadequate problem
specification produces false precision.
- Phase C2 (Divergent Candidate Generation) must precede Phase C3
because you cannot screen what has not been generated.
- Phase C3 (Hard-Constraint Screening) must precede Phase C4
because trade-off analysis among non-viable candidates is wasted
effort.
- Phase C4 (Trade-Off Analysis) must precede Phase C5 because
catastrophic-path testing is expensive and should only run on
survivors.
- Phase C5 (Catastrophic-Path Stress Test) must precede Phase C6
because selection without stress-test evidence is premature
convergence.
- Phase C6 (Selection and Rationale Capture) must precede Phase C7
because verification plans depend on a chosen solution.
- Phase C7 (Verification Plan and Handoff) is the final phase and
produces the artifact the implementation team will consume.
Never skip or reorder phases. If the user asks to skip a phase,
explain the dependency and ask if they want to proceed with reduced
rigor (and note this in the output).
Reverse-Contamination Guard — Continuous Application
Where the Problem-Space Exploration framework guards against
solution-contamination of the problem, this framework guards against
the opposite failure: solution decisions that silently redefine the
problem to make the solution look better.
At the end of every phase, re-read the output and flag any
statement that:
- Relaxes a mandatory property from the problem analysis without
going through an explicit re-analysis event.
- Reinterprets a hard constraint as a soft preference to preserve
a preferred candidate.
- Redefines the primary actor, trigger, or success condition from
Phases 1–2 of the problem analysis.
- Substitutes the narrative path from Phase 7 with a more
convenient one.
Flagged statements must be either reverted (the solution must adapt
to the problem, not the reverse) or escalated to the user as a
re-analysis event, which requires returning to the problem-space
framework.
Output Format
The final output is a single Markdown document with the following
structure:
---
title: "Solution-Space Convergence: [Subject]"
framework_version: "1.0.0"
problem_analysis_reference: "[Link or identifier of input PSE doc]"
problem_class: "[Class inherited from PSE]"
date: "[ISO date]"
disclaimer: >
No information within this document should be taken for granted.
Any statement or premise not backed by a real logical definition
or verifiable reference may be invalid, erroneous, or a
hallucination. This analysis is a reasoning instrument, not a
truth claim. The solution selected here is conditional on the
correctness of the input problem-space analysis.
---
# Solution-Space Convergence: [Subject]
## Input Reference
[Summary of the problem statement, mandatory properties, hard
constraints, and narrative paths inherited from the problem analysis.
This section is a restatement, not a revision.]
## Phase C0 — Reverse-Contamination Guard
[Detection criteria established for this convergence]
## Phase C1 — Input Adequacy Check
[Pass/Fail per required input element + gaps noted]
## Phase C2 — Divergent Candidate Generation
[≥3 structurally distinct candidates, each with architecture sketch,
mechanism summary, and assumed operating regime]
## Phase C3 — Hard-Constraint Screening
[Mandatory-property-by-candidate matrix: Pass / Fail / Uncertain,
with violation test result for each Fail]
## Phase C4 — Trade-Off Analysis (Soft Criteria)
[Pugh matrix or weighted decision matrix among C3 survivors,
with criteria traceable to the problem class and stakeholder
topology from the PSE input]
## Phase C5 — Catastrophic-Path Stress Test
[For each survivor, analysis against Phase 7 Path B from the PSE
input: dampens / neutral / amplifies the cascade]
## Phase C6 — Selection and Rationale Capture (QOC)
[Question / Options / Criteria notation for the primary selection
decision and any sub-decisions, with positive/negative assessments]
## Phase C7 — Verification Plan and Handoff
[Property → Mechanism → Test mapping, residual risks, and
implementation handoff checklist]
## Reverse-Contamination Audit Log
[Summary of all reverse-contamination flags raised during the
analysis and how they were handled]
## Re-Analysis Triggers
[Explicit list of conditions under which the convergence must be
re-run, inherited from the PSE transition triggers plus any
new ones surfaced during convergence]
Output Rules
- The disclaimer frontmatter is mandatory. Never omit it.
- Every phase heading must appear even if the content is thin.
An empty phase with a note ("Insufficient input to complete this
phase — requires [X]") is better than a missing phase.
- The Reverse-Contamination Audit Log at the end collects all flags
raised during the analysis.
- Use the vocabulary established in the PSE input's Phase 3
consistently. Do not introduce new terms for concepts already
named in the problem analysis; if a new term is needed, note it
explicitly and justify why.
Iteration Guidance
The framework expects multiple passes:
- First pass usually reveals that the candidate set is too narrow
(fewer than 3 structurally distinct options) or that several
candidates fail Phase C3 silently.
- Second pass usually reveals that soft criteria in Phase C4 are
weighted in a way that pre-determines the winner. Reweight and retest.
- Third pass should stabilize.
If the third pass does not stabilize, one of the following is true:
- The problem analysis is under-specified (return to PSE).
- The candidate set is missing a fundamentally different architecture
(return to C2 with a fresh perspective).
- The problem class is wrong (the weights that feel right do not match
the declared class — return to PSE Problem Class Declaration).
What This Skill Does NOT Do
- It does not replace the Problem-Space Exploration framework. It
consumes PSE output.
- It does not produce implementation code. The output of Phase C7 is a
specification with verification criteria, not an implementation.
- It does not make selection decisions for the user in Mode B. The
framework structures the decision and makes rationale explicit; the
user makes the call.
- It does not guarantee the chosen solution is correct. It guarantees
that the reasoning from problem to solution is auditable —
someone reading the output can reconstruct why this candidate was
chosen and what would falsify the choice.