スキル一覧に戻る
Rwb3n

ground-cycle

by Rwb3n

0🍴 0📅 2026年1月22日
GitHubで見るManusで実行

SKILL.md


name: ground-cycle description: HAIOS Ground Cycle for loading architectural context before cognitive work. Use before plan-authoring, investigation, or implementation cycles. generated: 2026-01-06 last_updated: '2026-01-06T21:07:54'

Ground Cycle

This skill defines the PROVENANCE-ARCHITECTURE-MEMORY-CONTEXT MAP cycle for loading context before any cognitive work begins.

When to Use

Automatically invoked at the start of:

  • plan-authoring-cycle (before AMBIGUITY phase)
  • investigation-cycle (before HYPOTHESIZE phase)
  • implementation-cycle (before PLAN phase)

Manual invocation: Skill(skill="ground-cycle", args="{work_id}")


The Cycle

PROVENANCE --> ARCHITECTURE --> MEMORY --> CONTEXT MAP --> [return GroundedContext]
     |               |              |              |
  Read WORK.md   Load epoch    Query prior     Build output
  Traverse      architecture   patterns       for caller
  spawned_by

1. PROVENANCE Phase

Goal: Traverse spawned_by chain to find source investigations.

Actions:

  1. Read work item: docs/work/active/{work_id}/WORK.md
  2. Extract spawned_by_investigation field
  3. If set, read source investigation's key findings
  4. Recursively traverse: E2-271 -> INV-057 -> INV-052
  5. Collect provenance chain

Exit Criteria:

  • Work item WORK.md read
  • spawned_by_investigation traversed (if set)
  • Provenance chain collected

Tools: Read, Glob


2. ARCHITECTURE Phase

Goal: Load epoch architecture for the work item.

Actions:

  1. Extract epoch field from work item (e.g., "E2")
  2. Load epoch definition: .claude/haios/epochs/{epoch}/EPOCH.md
  3. Load REQUIRED READING from epoch's architecture/
  4. Extract chapter and arc fields if present
  5. Load chapter/arc REQUIRED READING if applicable

Exit Criteria:

  • EPOCH.md loaded
  • Architecture documents loaded
  • REQUIRED READING section parsed

Tools: Read


3. MEMORY Phase

Goal: Query memory for relevant strategies and prior work.

Actions:

  1. Query memory for epoch architecture: "Epoch 2.2 architecture modules"
  2. Query memory for provenance chain topics
  3. Query memory for work item related concepts
  4. Collect relevant strategies

Exit Criteria:

  • Memory queried with epoch context
  • Strategies collected

Tools: mcp__haios-memory__memory_search_with_experience


4. CONTEXT MAP Phase

Goal: Build and output GroundedContext.

Actions:

  1. Assemble GroundedContext object:
    GroundedContext:
      epoch: E2
      chapter: C3-WorkInfra
      arc: ARC-001-ground-cycle
      provenance_chain: [E2-271, INV-057, INV-052]
      architectural_refs:
        - .claude/haios/epochs/E2/EPOCH.md
        - .claude/haios/epochs/E2/architecture/S17-modular-architecture.md
      memory_concepts: [80858, 80866, 80918]
      required_reading_loaded: true
    
  2. Return to calling cycle

Exit Criteria:

  • GroundedContext assembled
  • Returned to calling cycle

Tools: (internal)


GroundedContext Output Schema

GroundedContext:
  # Work item context
  epoch: str                      # E2
  chapter: str | null             # C3-WorkInfra
  arc: str | null                 # ARC-001-ground-cycle

  # Provenance
  provenance_chain: list[str]     # [E2-271, INV-057, INV-052]

  # Architectural documents loaded
  architectural_refs: list[str]   # File paths loaded

  # Memory context
  memory_concepts: list[int]      # Concept IDs from memory query

  # Status
  required_reading_loaded: bool   # True if REQUIRED READING parsed

Composition Map

PhasePrimary ToolMemory Integration
PROVENANCERead, Glob-
ARCHITECTURERead-
MEMORYmemory_search_with_experienceQuery epoch + provenance
CONTEXT MAP(internal)-

Integration Points

Calling Cycles

ground-cycle is invoked as "Phase 0" before cognitive work:

CycleWhen CalledBefore Phase
plan-authoring-cycleWork item enters plan nodeAMBIGUITY
investigation-cycleWork item enters discovery nodeHYPOTHESIZE
implementation-cycleWork item enters implement nodePLAN

Invocation Pattern

# In plan-authoring-cycle, before AMBIGUITY phase:
Skill(skill="ground-cycle", args="{work_id}")
# ground-cycle returns GroundedContext
# plan-authoring-cycle continues with architectural awareness

Call Chain Context

/new-plan E2-XXX (or /implement, /new-investigation)
    |
    +-> plan-authoring-cycle
            |
            +-> ground-cycle        # <-- THIS SKILL
            |       Input: work_id
            |       Output: GroundedContext
            |
            +-> AMBIGUITY phase (with grounded context)
            +-> ANALYZE phase
            +-> AUTHOR phase
            +-> VALIDATE phase

Edge Cases

CaseHandling
No spawned_by_investigationSkip PROVENANCE traversal, continue
No epoch field in work itemDefault to "E2" (current epoch)
Epoch directory missingWarn and continue without architecture
Memory MCP timeoutWarn and continue without strategies

Key Design Decisions

DecisionChoiceRationale
Separate skillNot inline codeReusable by multiple cycles
4 phasesPROVENANCE->ARCHITECTURE->MEMORY->CONTEXT MAPLogical progression
Return GroundedContextStructured outputDownstream cycles consume consistently
No state persistenceIn-memory onlyContext is session-scoped
Token budget awareReference S15GroundedContext fits within L4 budget

  • ARC-001-ground-cycle: Arc containing this and related work items
  • S17-modular-architecture.md: Module interfaces ground-cycle loads
  • S2C-work-item-directory.md: Portal system for provenance traversal
  • S14-bootstrap-architecture.md: Context loading hierarchy
  • S15-information-architecture.md: Token budgets
  • plan-authoring-cycle: Consumer - invokes ground-cycle before AMBIGUITY
  • investigation-cycle: Consumer - invokes ground-cycle before HYPOTHESIZE
  • implementation-cycle: Consumer - invokes ground-cycle before PLAN

Created Session 179, E2-276

スコア

総合スコア

50/100

リポジトリの品質指標に基づく評価

SKILL.md

SKILL.mdファイルが含まれている

+20
LICENSE

ライセンスが設定されている

0/10
説明文

100文字以上の説明がある

0/10
人気

GitHub Stars 100以上

0/15
最近の活動

3ヶ月以内に更新がある

0/10
フォーク

10回以上フォークされている

0/5
Issue管理

オープンIssueが50未満

+5
言語

プログラミング言語が設定されている

+5
タグ

1つ以上のタグが設定されている

0/5

レビュー

💬

レビュー機能は近日公開予定です