ai-engineering-from-scratch/learning-paths/agent-skills.json

{
  "schemaVersion": 1,
  "id": "agent-skills",
  "title": "Agent Skills Engineering",
  "summary": "Build, route, secure, evaluate, package, and verify Agent Skills in real hosts.",
  "estimatedMinutes": 570,
  "stateFile": "AGENT-SKILLS-LEARNING.md",
  "prerequisites": [
    {
      "type": "knowledge",
      "required": true,
      "title": "Files, Python, and the command line",
      "description": "You can edit a text file, inspect a directory, and run a Python command. No agent framework experience is required."
    },
    {
      "type": "software",
      "required": true,
      "title": "Node.js, npx, and Python 3 for real labs",
      "description": "The install and bundled-script checkpoints require working node, npx, and python3 commands. The website and docs remain a conceptual fallback when these commands are unavailable."
    },
    {
      "type": "host",
      "required": true,
      "title": "One selected skill-capable host and install scope",
      "description": "Choose a supported agent host plus a writable project or user skill scope before installation. Record the exact host, scope, and installed path; do not infer discovery from installer success."
    },
    {
      "type": "lesson",
      "required": false,
      "title": "The Tool Interface",
      "path": "phases/13-tools-and-protocols/01-the-tool-interface",
      "description": "Skim this first if tool schemas and side-effect boundaries are new to you."
    },
    {
      "type": "lesson",
      "required": false,
      "title": "Tool Schema Design",
      "path": "phases/13-tools-and-protocols/05-tool-schema-design",
      "description": "Skim this first if you have not designed typed tool contracts."
    },
    {
      "id": "tool-poisoning-and-untrusted-instructions",
      "type": "knowledge",
      "required": true,
      "title": "Tool poisoning and untrusted instructions",
      "description": "Before Lesson 26, explain why skill and tool metadata is untrusted input. Lesson 15 is an optional refresher outside this five-lesson route when this preflight is not yet confirmed."
    }
  ],
  "invocation": {
    "codex": "learn-agent-skills, or choose it from /skills",
    "claudeCode": "/learn-agent-skills",
    "portableFallback": "Use learn-agent-skills to start or resume the Agent Skills Engineering path."
  },
  "conceptualFallback": "Read the website or each lesson's docs/en.md manually. Mark real-host discovery, invocation, script, update, and uninstall evidence pending until Node.js, npx, Python 3, a compatible host, and a writable install scope are available.",
  "quickStart": {
    "estimatedMinutes": 10,
    "lessonPath": "phases/13-tools-and-protocols/22-skills-and-agent-sdks",
    "goal": "Create a small skill, install the complete reviewer bundle into a real host, invoke it explicitly, verify its output, and remove it cleanly.",
    "installCommand": "npx skills add rohitg00/ai-engineering-from-scratch --skill skill-contract-reviewer --full-depth",
    "workingDirectory": "An empty learner workspace. Save its absolute path as TARGET_ROOT before installation.",
    "expectedEvidence": [
      "The installer reports the selected host and destination.",
      "The host discovers skill-contract-reviewer after a rescan or new session.",
      "An explicit invocation returns a validation report and primitive choice, plus the resolved script path, target path, cwd, argv, and exit code.",
      "The uninstall command removes only skill-contract-reviewer from the selected host."
    ]
  },
  "lessons": [
    {
      "order": 1,
      "phase": 13,
      "lesson": 22,
      "title": "Agent Skills: Portable Contract and Runtime Boundary",
      "path": "phases/13-tools-and-protocols/22-skills-and-agent-sdks",
      "minutes": 90,
      "required": true,
      "checkpoint": "Create, install, invoke, verify, and remove one complete skill bundle in a real host."
    },
    {
      "order": 2,
      "phase": 13,
      "lesson": 24,
      "title": "Skill Discovery and Progressive Disclosure",
      "path": "phases/13-tools-and-protocols/24-skill-discovery-and-progressive-disclosure",
      "minutes": 105,
      "required": true,
      "checkpoint": "Trace discovery, catalog publication, activation, and branch-specific resource loading separately."
    },
    {
      "order": 3,
      "phase": 13,
      "lesson": 25,
      "title": "Skill Invocation and Routing",
      "path": "phases/13-tools-and-protocols/25-skill-invocation-and-routing",
      "minutes": 105,
      "required": true,
      "checkpoint": "Exercise explicit invocation, implicit selection, abstention, and a near-miss request."
    },
    {
      "order": 4,
      "phase": 13,
      "lesson": 26,
      "title": "Skill Permissions, Sandboxes, and Trust",
      "path": "phases/13-tools-and-protocols/26-skill-permissions-sandboxes-and-trust",
      "minutes": 120,
      "required": true,
      "prerequisitePaths": [
        "phases/13-tools-and-protocols/25-skill-invocation-and-routing"
      ],
      "prerequisiteChecks": [
        "tool-poisoning-and-untrusted-instructions"
      ],
      "checkpoint": "Distinguish instructions, permission prompts, sandbox containment, and independent verification with observed evidence."
    },
    {
      "order": 5,
      "phase": 13,
      "lesson": 27,
      "title": "Skill Evals, Packaging, and Portability",
      "path": "phases/13-tools-and-protocols/27-skill-evals-packaging-and-portability",
      "minutes": 150,
      "required": true,
      "checkpoint": "Run the release gate in a real host, record routing and authority evidence, and verify upgrade and uninstall behavior."
    }
  ],
  "optionalLessons": [
    {
      "phase": 13,
      "lesson": 23,
      "title": "Capstone: Stateless Tool Ecosystem",
      "path": "phases/13-tools-and-protocols/23-capstone-tool-ecosystem",
      "minutes": 120,
      "required": false,
      "entryRule": "Complete the Agent Skills route and Phase 13 lessons 06 through 20 before starting this systems integration capstone."
    }
  ]
}