hm-l3-integration-contracts

Files (14)

• SKILL.md
• .gitkeep · evals/
• evals.json · evals/
• .gitkeep · metrics/
• rich-gate-scorecard.md · metrics/
• .gitkeep · references/
• agent-to-skill-bindings.md · references/
• contract-schema.md · references/
• cross-lineage-rules.md · references/
• skill-to-agent-bindings.md · references/
• .gitkeep · scripts/
• validate-contracts.sh · scripts/
• .gitkeep · templates/
• .gitkeep · workflows/

SKILL.md (first lines)

## Overview

Define and enforce bidirectional agent↔skill bindings across all 5 Hivemind lineages. Every skill declares its intended agent audience; every agent domain declares required skill sets. This prevents orphan skills (no agent loads them), unnecessary cross-lineage loads (violating D-AD-01), and undocumented skill dependencies.

**This skill is the contract authority.** It is the source of truth for which agents load which skills and which skills are consumed by which agents.

## On Load

1. Read `references/agent-to-skill-bindings.md` — complete agent domain → required skills mapping
2. Read `references/skill-to-agent-bindings.md` — complete skill → consumed-by agents mapping  
3. Read `references/cross-lineage-rules.md` — D-AD-01 enforcement rules
4. Read `references/contract-schema.md` — the contract declaration schema for both SKILL.md and agent .md files
5. Validate the binding under question against the canonical tables

## Quick Jump

| Need | Section | Jump |
|------|---------|------|
| "Which skills does agent X load?" | Agent→Skill Table | [references/agent-to-skill-bindings.md](references/agent-to-skill-bindings.md) |
| "Which agents consume skill Y?" | Skill→Agent Table | [references/skill-to-agent-bindings.md](references/skill-to-agent-bindings.md) |
| "Can an hf agent load an hm skill?" | Cross-Lineage Rules | [references/cross-lineage-rules.md](references/cross-lineage-rules.md) |
| "How do I declare bindings?" | Contract Schema | [references/contract-schema.md](references/contract-schema.md) |
| "Is this skill orphaned?" | Orphan Detection | [Orphan Detection Protocol](#orphan-detection-protocol) |
| "Validate all bindings" | Validation Script | [scripts/validate-contracts.sh](scripts/validate-contracts.sh) |
| "What if a binding is wrong?" | Self-Correction | [Self-Correction](#self-correction) |

## Agent-to-Skill Bindings

Each agent domain loads a mandatory set of skills. These are non-negotiable — an agent without its required skills is under-equipped for its task domain.

### Research Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Researcher | hm-l3-detective, hm-l3-deep-research, hm-l3-tech-stack-ingest, hm-l3-research-chain |
| Scout | hm-l3-detective, hm-l3-tech-stack-ingest |
| Investigator | hm-l3-detective, hm-l2-debug, hm-l2-completion-looping |
| Synthesizer | hm-l3-synthesis, hm-l3-deep-research |
| Technician | hm-l3-tech-stack-ingest, hm-tech-context-compliance |

### Planning Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Planner | hm-l2-spec-driven-authoring, hm-l2-planning-persistence |
| Strategist | hm-l2-spec-driven-authoring, hm-l2-roadmap-maintainability, hm-l2-feature-ecosystem |
| Architect | hm-l2-refactor, hm-l2-cross-cutting-change |
| Ecologist | hm-l2-feature-ecosystem, hm-l2-roadmap-maintainability |

### Implementation Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Executor | hm-l2-phase-execution, hm-l2-cross-cutting-change, hm-l2-test-driven-execution |
| Builder (hm) | hm-l2-test-driven-execution, hm-l2-cross-cutting-change |
| Integrator | hm-l2-phase-execution, hm-l2-production-readiness |
| Operator | hm-l2-phase-execution, hm-l2-completion-looping |

### Quality Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Reviewer | hm-l2-test-driven-execution, gate-l3-evidence-truth, hm-l2-production-readiness |
| Auditor | hm-l2-gate-orchestrator, gate-l3-evidence-truth, gate-l3-spec-compliance |
| Validator | hm-l2-test-driven-execution, gate-l3-evidence-truth |
| Critic | gate-l3-evidence-truth, hm-l2-test-driven-execution |
| Assessor | hm-l2-production-readiness, gate-l3-evidence-truth |

### Debug Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Debugger | hm-l2-debug, hm-l2-completion-looping |
| Investigator | hm-l3-detective, hm-l2-debug, hm-l2-completion-looping |

### Orchestration Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Orchestrator (L0) | hm-l2-gate-orchestrator, hm-l2-coordinating-loop |
| Coordinator (L1) | hm-l2-coordinating-loop, hm-l2-subagent-delegation-patterns, hm-l2-gate-orchestrator |
| Guardian | hm-l2-phase-loop, hm-l2-completion-looping |
| Finisher | hm-l2-completion-looping, hm-l2-test-driven-execution |
| Persistor | hm-l2-planning-persistence, hm-l2-completion-looping |

### Meta-Builder Domain (hf-* lineage)
| Agent Type | Required Skills |
|-----------|----------------|
| Orchestrator (hf L0) | hf-l2-agent-composition, hf-l2-agents-and-subagents-dev, hf-l2-delegation-gates |
| Coordinator (hf) | hf-l2-agent-composition, hf-l2-agents-and-subagents-dev |
| Agent Builder | hf-l2-agent-composition, hf-l2-agents-and-subagents-dev, hf-l2-naming-syndicate |
| Skill Builder | hf-l2-skill-synthesis, hf-l2-use-authoring-skills, hf-l2-naming-syndicate |
| Command Builder | hf-l2-command-dev, hf-l2-command-parser, hf-l2-naming-syndicate |
| Tool Builder | hf-l2-custom-tools-dev, hf-l2-naming-syndicate |
| Auditor (hf) | hf-l2-auditor (self-referential), gate-l3-evidence-truth |
| Refactorer (hf) | hf-l2-use-authoring-skills, hm-l2-refactor |
| Synthesizer (hf) | hf-l2-skill-synthesis, hm-l3-synthesis |
| Prompter | hf-l2-agent-composition, hf-l2-context-absorb |

### Discovery Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Brainstormer | hm-l2-brainstorm |
| Mentor | hm-l2-brainstorm, hm-l2-product-validation |

### Writing Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Writer | hm-l2-spec-driven-authoring |
| Curator | hm-l2-production-readiness, hm-l2-roadmap-maintainability |

### Routing Domain
| Agent Type | Required Skills |
|-----------|----------------|
| Router | hm-l2-skill-router, hm-l2-lineage-router |
| Connector | hm-l2-cross-cutting-change, hm-l2-coordinating-loop |

## Skill-to-Agent Bindings

Each skill declares which agent types consume it. This enables orphan detection (skills with zero consumers) and overlap detection (same skill loaded by agents that don't need it).

### Research Pipeline Skills
| Skill | Consumed By | Lineage |
|-------|------------|---------|
| hm-l3-detective | hm-l2-researcher, hm-l2-scout, hm-l2-investigator | hm STRICT |

From shynlee04/hivemind-plugin, MIT.

More from shynlee04/hivemind-plugin