Compare candidates under declared parity and selection policy to produce a non-dominated set.
npx skills add https://github.com/m0n0x41d/haft --skill h-compare
Please help me install this Agent Skill. Skill: h-compare Source: https://github.com/m0n0x41d/haft/tree/main/internal/cli/skill/h-compare Command: npx skills add https://github.com/m0n0x41d/haft --skill h-compare
When several solution candidates already exist, teams often collapse the choice into a single hidden score or let presentation order decide. This Skill enforces a fair, explicit comparison: constraints, targets, parity conditions, and a predeclared selection policy are stated before scoring, so trade-offs stay visible and the final value judgment remains with the operator.
Ask the agent to compare your existing candidates, for example: compare Postgres, SQLite, and DuckDB for our analytics workload under equal hardware and a declared selection policy.
High-intent search queries and answers about installing and using this skill.
Declare the characteristic space first: separate hard constraints from optimization targets and passive observations. Then declare parity conditions and a selection policy before scoring, evaluate one dimension across all candidates, and report the non-dominated set with explicit trade-offs.
A non-dominated set contains the candidates that are not strictly worse than another option on every dimension. It preserves the real trade-off among survivors instead of collapsing the choice into one score or a single hidden winner.
Persist only when the operator explicitly asks to save, or when a concrete receiving use needs addressable replay. Ordinary comparisons stay conversational; persistence uses haft_problem characterize and haft_solution compare with exact record references.
No. It returns constraints, trade-offs, and a non-dominated set, leaving the value choice with the operator. A binding choice requires a direct unambiguous operator request routed through the h-decide skill.
Declaring the policy before seeing results prevents motivated reasoning where the policy is bent to favor a preferred outcome. Changing it afterward is allowed only when the policy change itself is made explicit.
kryptonum-devSeparate observed problems from stated causes and proposed directions before planning.
nearaiDetect explicit decisions in conversation and persist them as durable commitment records.
SpaceZephyrRetrieves and synthesizes personal memory records from L1-L4 with cited sources.
SPIRAL-EDWINCompute TOPSIS scores and rank alternatives from a decision matrix.
