Skip to content

putervision/behavior-mcp / .agents/skills/agent-reasoning-mcp

agent-reasoning-mcp

Verified

Teaches the agent to use the Strategic Agent Reasoning MCP server for BDI goals, utility scoring, risk evaluation, and replanning.

Strategic Agent Reasoning (agent-reasoning-mcp)

This skill provides step-by-step guidance and operational patterns for interacting with @putervision/agent-reasoning-mcp with project slug "behavior-mcp".


1. Role in the PuterVision Pentad

  • Workflow State (state-memory-mcp): Persistent task DAGs, decisions, milestones, and blockers.
  • Perception (vision-memory-mcp): Visual layout caching, screenshots, and visual specifications.
  • Spatial World (world-model-mcp): Persistent 3D/2D coordinates, bounding boxes, and topological relations.
  • Strategic Reasoning (agent-reasoning-mcp): BDI goal decomposition, multi-attribute expected utility calculation, belief decay, risk assessment, and replanning.
  • Tactical Execution (behavior-mcp): Deterministic ~60Hz browser behavior tree execution and reactive preemption.

2. Core Operational Sequence

  1. Initialize Objectives: Call set_goal with action: "create" to define top-level goals and action: "decompose" to establish subgoals.
  2. Configure Utility Profile: Tune agent priorities using set_utility_weights (aggression, caution, greed, exploration).
  3. Situational Trade-off Scoring: Call evaluate_situation with action: "snapshot" to rank candidate actions using Pareto utility theory.
  4. Intention Dispatch: Translate chosen action into an execution directive via manage_intentions.
  5. Reactive Replanning: If an unexpected obstacle or blocker emerges, invoke replan.

3. Complete 15 Consolidated MCP Tools Reference

Tool NameKey ActionsKey ParametersDescription
set_goalcreate, update, get, list, decompose, archivetitle, description, priority, parent_id, subgoalsHierarchical BDI goal management and task DAG decomposition.
evaluate_situationsnapshot, quicksnapshot, candidates, utility_profileMulti-attribute utility evaluation ranking candidate actions from environment state.
replanblocker, recovery, alternativegoal_id, blocker_description, strategyAdaptive DAG reconstruction and alternative path discovery upon obstacles.
assess_riskassess, matrixhazards, tolerance, mitigationsQuantitative threat matrix and probabilistic risk scoring.
query_knowledgesearch, lookup, heuristicsquery, category, tagsKnowledge retrieval of past decision heuristics and domain heuristics.
set_utility_weightsconfigure, get, list, profilename, weights (aggression, caution, greed, exploration)Utility weight tuning and personality profile management.
get_decision_traceget, list, explaintrace_id, limitExplainable chain-of-thought rationale playback and auditing.
manage_beliefsset, get, decay, listkey, value, confidence, decay_rateStructured belief state with temporal exponential confidence decay.
manage_intentionscreate, get, list, dispatch, cancelgoal_id, behavior_name, parametersExecution directives queue connecting strategic plans to runtime engines.
manage_reasoning_dbstats, audit, snapshot, restore, pruneaction, name, descriptionDatabase diagnostics, snapshots, and SHA-256 Merkle audit verification.
classifyevaluationcategory, input, taxonomy, state_packZero-LLM deterministic classification against hierarchical taxonomy (<2ms SLA).
ask_noulevaluationcondition, state_pack, thresholdFast binary (Yes/No/Abstain) heuristic gate evaluating conditions (<2ms SLA).
ask_choiceevaluationchoices, context, state_packDeterministic multi-alternative selection ranking candidate choices (<2ms SLA).
ask_scoreevaluationtarget, metric, scale, state_packHeuristic utility evaluation scoring target entities on a bounded scale (<2ms SLA).
gate_intentionevaluationproject, proposed_action, state_packFast-path safety & feasibility filter checking preconditions before execution (<1ms SLA).
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 468kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall