{
  "skill_id": "gf-ops-agent-timeout-policies",
  "name": "Agent Timeout Policy Pack",
  "evaluation": "free",
  "quality": "agent-decision-grade",
  "functional": true,
  "price_usd": 5.99,
  "category": "Async Ops",
  "tier": "standard",
  "kind": "service",
  "tags": [
    "x402",
    "goods-factory",
    "service",
    "functional",
    "sealed-runtime",
    "gf-complete"
  ],
  "summary": "Timeout, deadline, and cancellation policies for multi-step agents so hung tools fail closed instead of burning budget.",
  "when_to_use": [
    "Agents hang on slow tools and keep burning tokens or holding worker slots indefinitely.",
    "You need consistent deadlines across orchestration, tool calls, and user-facing async jobs.",
    "Partial failures leave orphaned subtasks with no cancellation path."
  ],
  "outline": [
    "Set hierarchy of clocks: user-visible SLA, orchestration wall clock, per-tool timeout, and network connect/read timeouts.",
    "Map each tool class (LLM, HTTP, DB, browser, payment check) to a default timeout and max-retry with total budget caps.",
    "Define cancellation trees: parent job cancel propagates to children; document what is safe to abort mid-flight.",
    "Specify failure surfaces: user message, retry eligibility, dead-letter queue, and ops alert thresholds for hang rates.",
    "Add soak tests: inject artificial latency and verify jobs fail closed within policy without cascading retries."
  ],
  "capabilities": [
    "Layered deadline model from user SLA down to socket timeouts",
    "Per-tool timeout and retry budget matrix for common agent tool classes",
    "Cancellation propagation rules for multi-step async graphs",
    "Hang-rate alerts and latency injection soak checklist"
  ],
  "artifact_outputs": [
    "agent-timeout-policy.md",
    "tool-timeout-matrix.yaml",
    "cancellation-tree.md"
  ],
  "decision_criteria": [
    "Outline steps are specific to this skill (not generic templates)",
    "Sealed pack exposes createSkill().run with outline-driven steps",
    "Artifacts are named concrete deliverables",
    "No revenue guarantees or banned claims",
    "Free eval path works without payment"
  ],
  "sealed_after_pay": [
    "Full src/index.js runtime (createSkill, run, plan)",
    "src/runbook.js with skill-specific steps",
    "src/tools.js registry for agent ToolRegistry",
    "WORKBOOK.md operator procedure",
    "templates/ checklist and artifact shells",
    "schemas/output.schema.json"
  ],
  "runtime": {
    "entry": "src/index.js",
    "export": "createSkill",
    "install": "node scripts/install-unlocked-pack.mjs unlock.json ./skills"
  },
  "challenge": "https://lvlltd.com/api/pay?skill=gf-ops-agent-timeout-policies",
  "sample": "https://lvlltd.com/skills/gf-ops-agent-timeout-policies/sample.md",
  "pack_meta": "https://lvlltd.com/api/packs?skill=gf-ops-agent-timeout-policies",
  "teaser": "https://lvlltd.com/sealed/gf-ops-agent-timeout-policies/",
  "proof": "https://lvlltd.com/api/proof",
  "standards": [
    "x402",
    "ERC-7857"
  ],
  "operator_notes": [
    "Confirm pay path with GET /api/pay before purchasing.",
    "Dry-run createSkill().run({ goal, dry_run: true }) after unlock.",
    "Persist payment.txHash with skill_id for re-redeem.",
    "Never commit private keys or OPS_SECRET into pack usage.",
    "Category focus: Async Ops. Kind: service.",
    "Claim scrub: no guaranteed ROI, no fake testimonials.",
    "If outline 404s on production, redeploy public/skills after completion pass."
  ]
}