skill-system-graph（技能系统图）

Skill System Graph

skill-system-graph is an AI-first navigation layer for project dependencies. It reads skill specs and v2 behavior specs, preserves declared relationship types, and syncs that graph into PostgreSQL for queryable navigation and dependency health checks.

This skill is SKILL.spec.yaml-first. It does not read SKILL.behavior.yaml for canonical graph construction.

Purpose

• Answer dependency questions quickly: what depends on what, and what changes may cascade.
• Provide CLI primitives where show reads the current spec-scan model and neighbors/path/impact query persisted edges.
• Keep behavior discovery fast via incremental refresh using source hashes.

Output Contract

Agents should run with --format json by default for stable automation.

• json output: machine-readable object, one record per command.
• text output: human-readable fallback only.

Core Operations

skill-system-graph exposes graph subcommands in scripts/.

graph show

Display all known nodes and edges from the current spec-scan model.

python3 "$(pwd)/skills/skill-system-graph/scripts/graph_cli.py" show [--skills-dir skills] [--check-deps] [--format json|text]

JSON shape (agent-facing):

{
  "status": "ok",
  "nodes": [{"skill_name": "...", "description": "...", "spec_path": "...", "operations_count": 0, "content_hash": "...", "stub": false}],
  "edges": [{"source": "...", "target": "...", "edge_type": "depends_on|delegates_to|reads|writes|..."}],
  "node_count": 0,
  "edge_count": 0
}

When --check-deps is used, the payload also includes dependency_report with missing and conflicting dependency diagnostics plus suggested fixes.

graph neighbors <skill_name>

List incoming and outgoing direct neighbors for a specific skill.

python3 "$(pwd)/skills/skill-system-graph/scripts/graph_cli.py" neighbors skill-system-router [--format json|text]

JSON shape:

{
  "skill": "skill-system-router",
  "outgoing": [{"skill_name": "skill-system-postgres", "edge_type": "depends_on"}],
  "incoming": [{"skill_name": "skill-system-tkt", "edge_type": "delegates_to"}],
  "status": "ok"
}

graph path <from_skill> <to_skill>

Find the shortest dependency path between two skills.

python3 "$(pwd)/skills/skill-system-graph/scripts/graph_cli.py" path skill-system-memory skill-system-postgres [--max-depth 10] [--format json|text]

JSON shape:

{
  "from_skill": "skill-system-memory",
  "to_skill": "skill-system-postgres",
  "path": ["skill-system-memory", "skill-system-router", "skill-system-postgres"],
  "found": true,
  "max_depth": 10
}

graph impact <skill_name>

Return transitive dependents (all skills that (transitively) depend on the given one).

python3 "$(pwd)/skills/skill-system-graph/scripts/graph_cli.py" impact skill-system-postgres [--max-depth 10] [--format json|text]

JSON shape:

{
  "skill": "skill-system-postgres",
  "impact": [
    {"impact_skill": "skill-system-memory", "depth": 1, "path": ["skill-system-memory", "skill-system-postgres"]},
    {"impact_skill": "skill-system-gui", "depth": 2, "path": ["skill-system-gui", "skill-system-memory", "skill-system-postgres"]}
  ],
  "impact_count": 2,
  "max_depth": 10,
  "status": "ok"
}

graph refresh

Rebuild and sync the graph from discovered specs in one command.

python3 "$(pwd)/skills/skill-system-graph/scripts/graph_cli.py" refresh [--skills-dir skills] [--force] [--format json|text]

JSON shape:

{
  "status": "ok",
  "parsed": 12,
  "inserted": 12,
  "updated": 0,
  "skipped": 0,
  "removed": 0
}

If refresh detects a dependency cycle, it fails closed and returns the full cycle path.

Internal helper operations

• parse-specs: scans SKILL.spec.yaml, builds canonical graph model, returns parse summary.
• sync-graph: persists graph model into PostgreSQL with hash-aware upserts.

The orchestrator (skill-system-router) is expected to use these primitives to keep data fresh before calls that require persistence guarantees.

Dependency Direction and Effect

This skill delegates to no further skill for execution (delegates_to: []) and depends on:

• skill-system-postgres for persisted graph storage
• skill-system-behavior for spec schema and behavior tooling context

Notes for AI Use

• Prefer --format json in automation scripts.
• Use text mode for quick operator inspection during debugging.
• Treat graph refresh as the authoritative source update step before long-running dependency calculations.

{
  "schema_version": "2.0",
  "id": "skill-system-graph",
  "version": "0.1.0",
  "capabilities": ["graph-navigation", "behavior-query", "dependency-analysis"],
  "effects": ["fs.read", "db.read", "db.write"],
  "operations": {
    "parse-specs": {
      "description": "Parse all skill SKILL.spec.yaml files into a normalized graph model.",
      "input": {
        "skills_dir": {"type": "string", "required": false, "description": "Root directory that contains skills/"},
        "include_invalid": {"type": "boolean", "required": false, "description": "Include malformed spec metadata in the parse report"}
      },
      "output": {
        "description": "Normalized graph model plus parse report.",
        "fields": {"graph": "object", "parse_report": "object"}
      },
      "entrypoints": {
        "agent": "Use scripts/graph_core.py parse helpers to scan SKILL.spec.yaml files only"
      }
    },
    "sync-graph": {
      "description": "Sync the normalized graph model into PostgreSQL graph tables.",
      "input": {
        "skills_dir": {"type": "string", "required": false, "description": "Root directory that contains skills/"},
        "force": {"type": "boolean", "required": false, "description": "Force rebuild ignoring cached hashes"}
      },
      "output": {
        "description": "Inserted, updated, skipped, and removed counters.",
        "fields": {"sync_result": "object"}
      },
      "entrypoints": {
        "agent": "Use scripts/graph_core.py sync helpers to upsert graph state into PostgreSQL"
      }
    },
    "show": {
      "description": "Display the full skill graph from the current spec-scan graph model.",
      "input": {
        "skills_dir": {"type": "string", "required": false, "description": "Root directory that contains skills/"},
        "format": {"type": "string", "required": false, "description": "Output format: json or text"},
        "check_deps": {"type": "boolean", "required": false, "description": "Include missing/conflicting dependency diagnostics"}
      },
      "output": {
        "description": "Full graph view with nodes and edges.",
        "fields": {"graph_view": "object"}
      },
      "entrypoints": {
        "unix": ["python3", "{skill_dir}/scripts/graph_cli.py", "show"],
        "windows": ["python", "{skill_dir}/scripts/graph_cli.py", "show"]
      }
    },
    "neighbors": {
      "description": "List direct neighbor skills for a target skill.",
      "input": {
        "skill_name": {"type": "string", "required": true, "description": "Skill id to inspect"},
        "format": {"type": "string", "required": false, "description": "Output format: json or text"}
      },
      "output": {
        "description": "Outgoing and incoming neighbors with edge types.",
        "fields": {"neighbors": "object"}
      },
      "entrypoints": {
        "unix": ["python3", "{skill_dir}/scripts/graph_cli.py", "neighbors", "{skill_name}"],
        "windows": ["python", "{skill_dir}/scripts/graph_cli.py", "neighbors", "{skill_name}"]
      }
    },
    "path": {
      "description": "Find the shortest dependency path between two skills.",
      "input": {
        "from_skill": {"type": "string", "required": true, "description": "Source skill id"},
        "to_skill": {"type": "string", "required": true, "description": "Destination skill id"},
        "max_depth": {"type": "number", "required": false, "description": "Traversal depth guard"},
        "format": {"type": "string", "required": false, "description": "Output format: json or text"}
      },
      "output": {
        "description": "Shortest path list and found flag.",
        "fields": {"path_result": "object"}
      },
      "entrypoints": {
        "unix": ["python3", "{skill_dir}/scripts/graph_cli.py", "path", "{from_skill}", "{to_skill}"],
        "windows": ["python", "{skill_dir}/scripts/graph_cli.py", "path", "{from_skill}", "{to_skill}"]
      }
    },
    "impact": {
      "description": "Find all transitive dependents for one skill.",
      "input": {
        "skill_name": {"type": "string", "required": true, "description": "Skill id whose dependents are requested"},
        "max_depth": {"type": "number", "required": false, "description": "Traversal depth guard"},
        "format": {"type": "string", "required": false, "description": "Output format: json or text"}
      },
      "output": {
        "description": "Transitive dependents with count plus depth/path metadata.",
        "fields": {"impact_result": "object"}
      },
      "entrypoints": {
        "unix": ["python3", "{skill_dir}/scripts/graph_cli.py", "impact", "{skill_name}"],
        "windows": ["python", "{skill_dir}/scripts/graph_cli.py", "impact", "{skill_name}"]
      }
    },
    "refresh": {
      "description": "Rebuild graph from specs and refresh persistence in one command.",
      "input": {
        "skills_dir": {"type": "string", "required": false, "description": "Root directory that contains skills/"},
        "force": {"type": "boolean", "required": false, "description": "Force rebuild ignoring cached hashes"},
        "format": {"type": "string", "required": false, "description": "Output format: json or text"}
      },
      "output": {
        "description": "Parsed spec count plus sync counters.",
        "fields": {"refresh_result": "object"}
      },
      "entrypoints": {
        "unix": ["python3", "{skill_dir}/scripts/graph_cli.py", "refresh"],
        "windows": ["python", "{skill_dir}/scripts/graph_cli.py", "refresh"]
      }
    }
  },
  "stdout_contract": {
    "last_line_json": true,
    "note": "CLI commands support stable JSON output for automation. Internal helper operations are agent-executed."
  }
}