Skill

Bottomup topdown elicitation

Elicit a system's requirements both top-down (general) and bottom-up (concrete instances as probes), up/down-chunking, draining when K probes add zero

Primitives inside (11)

• coverage-map-convergence-check failsafe

  Verify convergence against an explicit coverage inventory (regions x types: probed/unprobed/excluded) -- 'I am out of probes' is not 'the system is drained'.

  When: About to declare an elicitation/discovery loop converged.

  Register interest — 1 credit (≈ €0.10) — no payment

• diverse-probe-sampling discipline

  Deliberately sample under-covered regions and the RAREST instance types as probes -- a drain counter is only trustworthy if the probes are diverse.

  When: Choosing which instances to probe next in a discovery/elicitation drain.

  Register interest — 1 credit (≈ €0.10) — no payment

• dual-direction-elicitation discipline

  Elicit requirements in BOTH directions -- top-down first-principles enumeration AND bottom-up real-instance probes -- each direction is structurally blind to what the other catches.

  When: Deriving requirements/specs for a general system from concrete material and/or stakeholder intent.

  Register interest — 1 credit (≈ €0.10) — no payment

• expanding-probe-list discipline

  Derive the NEXT round's probes from what the previous round learned (leads), not from a fixed up-front list -- a static probe list exhausts itself without exhausting the system.

  When: Planning round N+1 of an iterative discovery loop.

  Register interest — 1 credit (≈ €0.10) — no payment

• generalization-second-clean-round calibration

  Convergence needs two consecutive clean rounds where the SECOND is a generalization test on a fresh un-tuned slice -- a clean round on tuned material proves nothing.

  When: An iterative improve-and-verify loop is about to be declared converged.

  Register interest — 1 credit (≈ €0.10) — no payment

• k-consecutive-zero-drain-stop calibration

  Stop a discovery/elicitation drain only after K CONSECUTIVE zero-yield probes -- never after one zero and never at a fixed batch count; non-monotone bumps are normal.

  When: Deciding when an iterative discovery/elicitation loop is finished.

  Register interest — 1 credit (≈ €0.10) — no payment

• probe-full-depth-not-names gotcha-fix

  Probe the FULL instance (body, content, acceptance, relations, internal structure) — never its name, label or title: name-level classification yields all-confirmations and a FAKE early saturation that ends a discovery sweep or drain loop far too early (measured: name-matching 'saturated' at ~200 of 47k nodes while full-depth reads still yielded ~3 new design elements per node). When the finding-rate drops to ~0, first verify you are reading complete bodies before believing the corpus is dry.

  When: Any discovery/classification sweep or drain/saturation loop over a corpus where a finding-rate per item examined drives the stop decision.

  Register interest — 1 credit (≈ €0.10) — no payment

• real-instances-as-probes discipline

  Bottom-up discovery probes must be REAL instances (real data, real nodes, real cases) -- an invented probe can only confirm what you already believe.

  When: Selecting probe material for a bottom-up elicitation/discovery pass.

  Register interest — 1 credit (≈ €0.10) — no payment

• record-negative-results discipline

  A candidate requirement tested and REJECTED bounds the system -- record it as a finding; it is progress, not noise.

  When: A probe or test refutes a candidate requirement/hypothesis during elicitation.

  Register interest — 1 credit (≈ €0.10) — no payment

• scenario-boundary-probes discipline

  Stress-test already-elicited requirements with invented EDGE-CASE scenarios that force precision about concept boundaries -- distinct from discovery probes, which must be real.

  When: A requirements set exists and its concept boundaries (what falls under which requirement) are still fuzzy.

  Register interest — 1 credit (≈ €0.10) — no payment

• write-down-immediately discipline

  Capture each elicited item the moment it crystallises -- batching the write-up loses items and corrupts the zero-yield count the drain rule depends on.

  When: Findings are emerging continuously during a probe round or interview.

  Register interest — 1 credit (≈ €0.10) — no payment

Neighbour skills

Skills whose primitives are closest to this one (bge-m3 similarity):

• Agentic design pattern selector
• Context pipeline assembly
• LightRAG local ollama graphrag