yogsoth-ai

Present the GoalTree to the user for confirmation. Ask about reasonableness, missing elements, and priority ordering among sub-goals.

Reads full web pages for non-academic perspectives like blogs, tech reports, product pages, and industry analysis. A subagent reads pages in isolated contexts, with a hard constraint of at least 30 full web pages read.

Focuses within the user's chosen field(s), identifying specific sub-directions through deep paper and web research, then presenting ranked candidates. This skill is used after landscape-reconnaissance has identified fields of interest.

Structure the user's chosen direction into a formal goal tree using KAOS-style AND/OR decomposition. Validate feasibility against ActorProfile and ObstacleReport. Use after obstacle-analysis confirms the direction is viable.

Broad, shallow exploration of candidate research fields to understand available options before narrowing down. Useful for users discovering fields, especially in cold-start and warm-start scenarios.

Goal-Driven Requirement Refinement Engine for Research. Crystallize a user's fuzzy research intent into a North Star statement and structured ResearchBrief through adaptive dialogue and on-demand investigation.

Evaluates each candidate research field based on maturity, competition, entry barrier, and publication opportunity. Synthesizes broad web search results into a structured FieldPanorama, considering both niche approaches and direct frontal competition in hot fields.

Identify what blocks the user from pursuing their chosen direction, assess severity, propose mitigations with search-validated evidence, and get user acceptance. Use after direction-narrowing has identified a specific direction.

A simplified crystallization strategy for users with a general research direction but lacking specificity. It streamlines actor profiling and landscape reconnaissance, guiding through direction narrowing, obstacle analysis, goal decomposition, and north-star synthesis, ideal for when a user's first message reveals a general area but not a specific problem.

Understand the user's background, resources, constraints, and deep motivations to produce an ActorProfile that informs all downstream decisions. This tactic is used at the start of any crystallization process to model the user's capabilities, limitations, and intent.

Deep WHY probing, inspired by i* Intentionality modeling, aims to understand the user's motivation, success definition, risk tolerance, and other key preferences. This understanding of WHY is the most important SOP in actor-profiling, driving all subsequent actions.

Present obstacles, including their severity assessments and proposed mitigations, to the user. Ask if these obstacles are acceptable, and if not after two rounds, return to presenting candidates.

Rate each identified obstacle's difficulty — overcomability, time cost, workaround existence. Optionally, use search tools to validate assessments.

Fuse the GoalTree root node and user motivation into a single crystallized North Star statement.

Understand the user's available research resources, including compute, timeline, collaboration, data access, and experimental environment. 'TBD' is an acceptable answer for any item.

A comprehensive crystallization strategy for users with no research direction, covering actor profiling, landscape reconnaissance, direction narrowing, obstacle analysis, goal decomposition, and north-star synthesis. Use when the user's initial message reveals zero specificity about their research intent.

Understands the user's comprehensive background, including technical stack, project and research experience, publications, and research directions. It allows users to express interests beyond their resume and executes only once, never re-running.

Cross-reference the GoalTree against ActorProfile (capabilities), ObstacleReport (known barriers), and timeline (deadline feasibility). Identify infeasible paths and suggest OR alternatives.

Self-review the North Star + ResearchBrief for completeness, consistency, and clarity. If issues are found, return to the specific tactic/SOP for a targeted fix; otherwise, present the final output to the user for confirmation.

Enumerate barriers to pursuing the chosen research direction — knowledge, resource, capability, and competition barriers. Optionally, search tools may be used to discover obstacles not mentioned by the user.

Converges all accumulated context into a crystallized North Star statement and structured ResearchBrief, performing self-review before presenting to the user. Use as the final tactic in any start mode, where everything comes together.

Analyzes sub-directions within the user's chosen field, presenting ranked candidates. This SOP integrates sub-direction identification, skill-gap matching, and presentation, with depth scaling by start mode: cold-start for broad sub-directions, warm-start for specific sub-problems, and hot-start for granular technical details.

Propose concrete mitigation strategies for severe obstacles. Validate proposed mitigations using search tools to ensure realism and feasibility, avoiding theoretical assumptions.

Performs a paper landscape scan within selected field(s), with strict import of the literature-engine/literature-overview skill.

Propose 3-8 candidate research fields based on the full ActorProfile. When the user wants to explore beyond their current stack, use other ActorProfile signals (intentionality, boundary) to determine the exploration space, allowing for free exploration within the boundary.

Aggregate all accumulated context from the crystallization process into a structured ResearchBrief document. This is the final output artifact alongside the North Star — a comprehensive requirement context document for downstream research strategies.

Understand the strict boundaries of the user's research, including target venues, methodology preferences, areas to avoid, and advisor/team requirements. This applies to any research domain, not just ML/AI.

KAOS-style recursive goal decomposition, using AND decomposition for sub-goals that must all be satisfied, and OR decomposition for alternative paths where any one suffices. This process produces a GoalTree (DAG structure).

Quick web scanning for field landscape understanding. This skill strictly imports web-browsing/web-search capabilities, with a hard constraint of 10 brave_web_search calls per instance and at least 150 total search results before completion.

Express the user's chosen research direction as a formal goal statement in the format: 'Achieve [what], such that [effect], under [constraints]'. Confirm with user before proceeding to decomposition.

Present the field panorama to the user and gather their preferences — which fields interest them, which they reject, and why. A dialogue SOP that bridges landscape-synthesis output to user decision.

Conducts a quality check on GoalTree leaf nodes, verifying each is specific, actionable, and that the set fully covers the top goal, flagging issues for further decomposition.