Skill-Agent Mapping

Central registry documenting the ownership and consumption relationships between skills and agents in the multi-agent orchestration system.

Overview

Skills are domain expertise modules that define behavioral patterns and best practices. Each skill has:

• Owner Agent: The agent responsible for maintaining and embodying the skill
• Consumer Agents: Agents that reference the skill for guidance or constraints

Quick Reference Table

Visual Mapping

                    +----------------------------------------------------------+
                    |              agent-behavior-constraints                   |
                    |                   (Owner: System)                         |
                    |              Consumed by: ALL AGENTS                      |
                    +----------------------------------------------------------+
                                              |
            +---------------------------------+---------------------------------+
            |                                 |                                 |
            v                                 v                                 v
+-----------------------+       +-----------------------+       +-----------------------+
|        Riko           |       |        Senku          |       |       Alphonse        |
|   (Explorer Agent)    |       |   (Planner Agent)     |       |   (Verifier Agent)    |
+-----------------------+       +-----------------------+       +-----------------------+
| OWNS:                 |       | OWNS:                 |       | OWNS:                 |
| - exploration-strategy|       | - task-classification |       | - verification-gates  |
| - graphify-usage      |       | - prompt-refinement   |       |                       |
| - personal-kb-usage   |       |                       |       |                       |
| - agentsview-usage    |       |                       |       |                       |
+-----------------------+       +-----------------------+       +-----------------------+
| CONSUMES:             |       | CONSUMES:             |       | CONSUMES:             |
| - agent-behavior-     |       | - agent-behavior-     |       | - agent-behavior-     |
|   constraints         |       |   constraints         |       |   constraints         |
| - task-classification |       | - exploration-strategy|       | - evidence-driven-work|
|                       |       | - graphify-usage      |       |                       |
|                       |       | - personal-kb-usage   |       |                       |
|                       |       | - agentsview-usage    |       |                       |
+-----------------------+       +-----------------------+       +-----------------------+
            |                                 |                                 |
            |                                 |                                 |
            v                                 v                                 v
+-----------------------+       +-----------------------+       +-----------------------+
|        Loid           |       |       Lawliet         |       |     Orchestrator      |
|  (Executor Agent)     |       |  (Reviewer Agent)     |       |   (Coordination)      |
+-----------------------+       +-----------------------+       +-----------------------+
| OWNS: (none)          |       | OWNS: (none)          |       | OWNS: (none)          |
+-----------------------+       +-----------------------+       +-----------------------+
| CONSUMES:             |       | CONSUMES:             |       | CONSUMES:             |
| - agent-behavior-     |       | - agent-behavior-     |       | - task-classification |
|   constraints         |       |   constraints         |       | - prompt-refinement   |
| - verification-gates  |       | - verification-gates  |       | - evidence-driven-work|
| - exploration-strategy|       | - graphify-usage      |       |   (Phase 6:           |
|                       |       | - personal-kb-usage   |       |   expression.md)      |
|                       |       | - agentsview-usage    |       |                       |
|                       |       | - evidence-driven-work|       |                       |
+-----------------------+       +-----------------------+       +-----------------------+

+-----------------------+
|      Speedwagon       |
|  (Authoring Agent)    |
+-----------------------+
| OWNS: (none)          |
+-----------------------+
| CONSUMES:             |
| - agent-behavior-     |
|   constraints         |
| - exploration-strategy|
| - explainer-design-system  |
+-----------------------+

Skill Descriptions

exploration-strategy

Owner: Riko (Explorer Agent) Consumers: Senku, Loid, Speedwagon Location: skills/exploration-strategy/SKILL.md

Defines patterns for efficient codebase exploration including:

• Search tool selection (Glob, Grep, Read)
• Parallel search strategies
• Convergence criteria for stopping exploration
• Context gathering techniques

task-classification

Owner: Senku (Planner Agent) Consumers: Riko, Orchestrator Location: skills/task-classification/SKILL.md

Provides guidance on:

• Task complexity assessment (Trivial, Exploratory, Implementation, Complex, Research)
• Agent routing decisions
• Verification requirement determination
• Risk amplifier identification

prompt-refinement

Owner: Senku (Planner Agent) Consumers: Orchestrator Location: skills/prompt-refinement/SKILL.md

Handles:

• Ambiguity detection in user requests
• Clarification question strategies
• Structured task specification templates
• Orchestration eligibility detection

verification-gates

Owner: Alphonse (Verifier Agent) Consumers: Loid, Lawliet Location: skills/verification-gates/SKILL.md

Defines:

• Pre-commit and pre-complete gate types
• Verification commands by language/framework
• Failure handling protocols
• Override rules and documentation requirements

agent-behavior-constraints

Owner: System Consumers: All agents Location: skills/agent-behavior-constraints/SKILL.md

Establishes:

• Model routing rules (Opus vs Sonnet)
• Tool access permissions per agent
• Universal behavioral guardrails
• MCP tool preferences

graphify-usage

Owner: Riko (Explorer Agent) Consumers: Senku, Lawliet Location: skills/graphify-usage/SKILL.md

Covers:

• When to query the knowledge graph vs. use Grep/Read
• Tool selection decision table (query_graph, get_node, get_neighbors, get_community, god_nodes, graph_stats, shortest_path)
• Token hygiene rules (top_k/top_n limits, no raw JSON downstream, source_location citations)
• Result interpretation including confidence tags (EXTRACTED/INFERRED/AMBIGUOUS) and staleness discipline
• Standard query patterns (orientation, blast-radius, boundary verification)

personal-kb-usage

Owner: Riko (Explorer Agent) Consumers: Senku, Lawliet Location: skills/personal-kb-usage/SKILL.md

Covers:

• When to query the personal KB vs. the project graph vs. Grep/Read
• Tool selection decision table using mcp__personal-kb__* prefix (query_graph, get_node, get_neighbors, get_community, god_nodes, graph_stats, shortest_path)
• Token hygiene rules (top_k/top_n limits, no raw JSON downstream, absolute source_location citations)
• Result interpretation including confidence tags and privacy/authority caveats (personal KB is prior experience, not authoritative project docs)
• Cross-project recall patterns (prior decisions, anti-patterns, style preferences)

agentsview-usage

Owner: Riko (Explorer Agent) Consumers: Senku, Lawliet Location: skills/agentsview-usage/SKILL.md

Covers:

• When to search prior session history vs. the project graph vs. Grep/Read
• Tool selection decision table using mcp__plugin_agent-flow_agentsview__* prefix (search_sessions, list_sessions, get_session_overview, get_messages, search_content)
• Token hygiene rules (always paginate get_messages, prefer overview/search over transcript dumps)
• Precedent interpretation: history is a prior not a mandate, flag precedent-vs-current-requirement conflicts, trust-but-verify

evidence-driven-work

Owner: System (cross-cutting) Consumers: Lawliet, Alphonse (preloaded, with a "Running as a subagent" section that keeps their verdict formats intact); the Orchestrator reads references/expression.md in Phase 6; Riko reaches references/perception.md through exploration-strategy; auto-triggers in plain sessions Location: skills/evidence-driven-work/SKILL.md

Covers:

• Perception: find the observation surfaces a domain offers (effective state, plans, metrics, history, API behind a UI) and recommend equivalent tools to the user
• Recovery from wrong hypotheses, change discipline, target-environment verification habits
• Evidence standards and argument structure (finding, mechanism, evidence, impact, options, recommendation)
• Human-facing expression: outcome first, full narrative (background, change, purpose, impact), terms explained, clarity over decoration
• Defers to exploration-strategy, verification-gates, agent-behavior-constraints and agentsview-usage for what they own

Reference Files:

• references/perception.md, references/recovery-and-change.md, references/verification.md
• references/evidence-and-argument.md, references/expression.md

explainer-design-system

Owner: External — vendored (see source: in skills/explainer-design-system/SKILL.md) Consumers: Speedwagon Location: skills/explainer-design-system/SKILL.md (plus references/ directory)

Covers:

• Content philosophy for teaching code to non-technical learners (metaphor rules, tone, visual density)
• Design system (warm palette, typography, spacing tokens) used by /explain fragments
• HTML patterns for interactive elements: code↔English translators, chat animations, flow animations, quizzes, callouts, glossary tooltips
• Gotchas checklist to run before declaring a module done

Adapter note: the upstream skill targets a multi-module course directory; here it is narrowed to the single-module /explain pipeline (see the "Adapter note" at the top of the vendored SKILL.md).

Ownership Principles

1. Single Ownership: Each skill has exactly one owner agent (or System for cross-cutting concerns)
2. Clear Boundaries: Owners maintain and evolve the skill; consumers reference but don't modify
3. Explicit Dependencies: Consumer relationships are declared in both skill and agent frontmatter
4. System Skills: Skills owned by "System" apply universally and are not agent-specific

Consumption Guidelines

When an agent consumes a skill:

1. Reference the skill when making decisions in the skill's domain
2. Follow the patterns defined in the skill documentation
3. Defer to the owner for ambiguous interpretations
4. Report conflicts if skill guidance contradicts task requirements

Related Files

• Agent Definitions - Agent .md files with skill references
• Skill Directories - Individual skill documentation
• README - Plugin overview and architecture