Generates 3-5 distinct candidate approaches with weakest links for a current engineering question.
npx skills add https://github.com/m0n0x41d/haft --skill h-explore
Please help me install this Agent Skill. Skill: h-explore Source: https://github.com/m0n0x41d/haft/tree/main/internal/cli/skill/h-explore Command: npx skills add https://github.com/m0n0x41d/haft --skill h-explore
When tackling an engineering problem, teams often fixate on the first plausible approach and stop searching. This Skill forces divergent thinking by generating 3-5 genuinely distinct candidate approaches for a current question, keeping each option's weakest link and evidence gap visible so premature convergence is avoided.
Ask the agent to use h-explore to generate distinct candidate approaches for your current design question, keeping each option's weakest link visible.
High-intent search queries and answers about installing and using this skill.
Invoke h-explore with your current question, cue, or problem reference. It produces 3-5 candidates that differ in kind rather than degree, each with a mechanism, expected benefit, weakest link, and evidence gap, returned conversationally by default.
Use h-explore when the current need is generating options, not diagnosing a failure or comparing already available alternatives. Comparison, decision, planning, and implementation are separate later steps, not automatic follow-ups.
No. Candidates stay conversational unless there is explicit save intent or a concrete reliance-bearing receiving use. Only then does it persist one Note per candidate and a SolutionPortfolio through the Haft MCP tools.
Unavailable project basis, known absence, or explicit abstention is non-blocking. The exploration continues without inventing profiles, entities, or artifacts, and no typed memory is persisted merely because a read failed.
Candidate generation is not a recommendation, choice, authorization, or work plan. It preserves uncertainty and stops once the option field is adequate; it does not select, rank for decision, or implement any variant.
zrosenfieldExtract structured Decision Records from video or audio transcripts.
vincent-kkGenerate and evaluate multiple interpretations of a single requirement.
synaptiaiRetrieve prior decisions and patterns from session, project, or global memories.
robkamDiagnose QA gate failures and implement root-cause fixes with rerun plans.
