Add page-level documentation evaluations

[thodges-gh]

Description

• Add strict, schema-backed page evaluation metadata and enroll three documentation pages for human and agent use.
• Add inventory, skill-filtered inventory, changed-page inspection, and run commands. Bind the first runnable evaluation to the CRE Read Data Feeds page and the registered chainlink-cre-skill.
• Generate and grade one response against the page's ordered outcomes, or run the offline known-bad fixture with zero provider calls. Store local results under ignored evals/runs/.
• Pin Promptfoo to 0.122.0 and raise the Node.js floor to 22.22.0.

Verification

• Promptfoo resolved and is pinned to 0.122.0; the Node.js floor is 22.22.0.
• Full Jest: 11 suites and 244 tests passed, exit 0.
• Typecheck exited 0.
• Full lint exited 0 with 0 errors and 111 existing warnings.
• Format check exited 0.
• Focused page evaluator: 86 tests passed, exit 0.
• Inventory, skill-filtered inventory, and changed-page demos each exited 0.
• Final known-bad smoke exited 0 and reported FAIL cre-read-data-feeds 2/3; its ignored JSON was mode 0600 and recorded providerCalls: 0.
• Independent parser, output, and path reviews passed with no high- or medium-severity findings remaining.

npm run page-evaluation -- run --eval cre-read-data-feeds --skills-repo ../chainlink-agent-skills

Review the ignored evals/runs/*.json result.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
documentation	Ready	Preview	Aug 26, 2026 10:45pm

[thodges-gh]

Output:

{
  "schemaVersion": 1,
  "runId": "20260826T223422Z-cre-read-data-feeds-fde085a6-8c12-4c62-883c-28791c3e74c6",
  "kind": "model",
  "evalId": "cre-read-data-feeds",
  "page": "src/content/cre-templates/read-data-feeds.mdx",
  "skills": [
    "chainlink-cre-skill"
  ],
  "usableBy": [
    "human",
    "agent"
  ],
  "outcomes": [
    "The response recommends regular Data Feeds for single-value feeds and MVR Data Feeds for bundled multi-value feeds.",
    "The response lists Go and TypeScript choices with the exact regular IDs read-data-feeds-go and read-data-feeds-ts and the exact MVR IDs read-mvr-data-feeds-go and read-mvr-data-feeds-ts.",
    "The response explains that the workflow runs every 10 minutes on a cron trigger, uses the CRE chain reader to call the feed contract, and returns scaled regular-feed values or decoded MVR fields."
  ],
  "providers": {
    "generator": "openai:gpt-5.6-terra",
    "grader": "openai:gpt-5.6-sol"
  },
  "response": "Choose based on the shape of data you need:\n\n- **Regular Data Feeds**: use these for a conventional **single-value feed**, such as BTC/USD or ETH/USD. The workflow reads `decimals()` and `latestAnswer()` from each feed contract, scales the answer using its decimals, logs it, and returns a JSON array of feed results.\n- **MVR (Multiple Variable Response) Data Feeds**: use these when one feed supplies a **bundle of multiple related values**. The workflow reads `latestBundle()` and `bundleDecimals()` from a `BundleAggregatorProxy`, decodes the returned bytes into typed fields, scales numeric fields as appropriate, and returns a structured JSON object. The provided MVR example uses the S&P Global SSA EURC feed on Base.\n\n## Template choices\n\n| Use case | Language | CRE CLI template ID |\n|---|---|---|\n| Regular single-value feeds | TypeScript | `read-data-feeds-ts` |\n| Regular single-value feeds | Go | `read-data-feeds-go` |\n| MVR bundled multi-value feeds | TypeScript | `read-mvr-data-feeds-ts` |\n| MVR bundled multi-value feeds | Go | `read-mvr-data-feeds-go` |\n\n## How the cron-based chain-reader flow works\n\n1. A **cron trigger fires every 10 minutes**.\n2. The CRE workflow uses the **CRE chain reader** to make read-only contract calls on the configured EVM network.\n3. For a regular feed, it calls:\n   - `decimals()` to learn the feed’s precision.\n   - `latestAnswer()` to obtain the current integer answer.\n4. The workflow applies the decimals to present the regular-feed answer as a scaled value, then returns the resulting feed entries as JSON.\n5. For an MVR feed, it calls:\n   - `latestBundle()` to retrieve the encoded bundle.\n   - `bundleDecimals()` to obtain precision metadata for bundle fields.\n6. It decodes the bundle bytes according to the feed’s field schema, scales numerical values, and returns the decoded fields as a structured JSON result.\n\nIn short: select **regular** templates for one price/value per feed contract; select **MVR** templates when the feed is intentionally published as a typed, multi-field bundle.",
  "outcomeResults": [
    {
      "outcome": "The response recommends regular Data Feeds for single-value feeds and MVR Data Feeds for bundled multi-value feeds.",
      "pass": true,
      "evidence": "In short: select **regular** templates for one price/value per feed contract; select **MVR** templates when the feed is intentionally published as a typed, multi-field bundle.",
      "reason": "The response clearly recommends each feed type according to whether the data is single-value or bundled multi-value."
    },
    {
      "outcome": "The response lists Go and TypeScript choices with the exact regular IDs read-data-feeds-go and read-data-feeds-ts and the exact MVR IDs read-mvr-data-feeds-go and read-mvr-data-feeds-ts.",
      "pass": true,
      "evidence": "Choose based on the shape of data you need:\n\n- **Regular Data Feeds**: use these for a conventional **single-value feed**, such as BTC/USD or ETH/USD. The workflow reads `decimals()` and `latestAnswer()` from each feed contract, scales the answer using its decimals, logs it, and returns a JSON array of feed results.\n- **MVR (Multiple Variable Response) Data Feeds**: use these when one feed supplies a **bundle of multiple related values**. The workflow reads `latestBundle()` and `bundleDecimals()` from a `BundleAggregatorProxy`, decodes the returned bytes into typed fields, scales numeric fields as appropriate, and returns a structured JSON object. The provided MVR example uses the S&P Global SSA EURC feed on Base.\n\n## Template choices\n\n| Use case | Language | CRE CLI template ID |\n|---|---|---|\n| Regular single-value feeds | TypeScript | `read-data-feeds-ts` |\n| Regular single-value feeds | Go | `read-data-feeds-go` |\n| MVR bundled multi-value feeds | TypeScript | `read-mvr-data-feeds-ts` |\n| MVR bundled multi-value feeds | Go | `read-mvr-data-feeds-go` |\n\n## How the cron-based chain-reader flow works\n\n1. A **cron trigger fires every 10 minutes**.\n2. The CRE workflow uses the **CRE chain reader** to make read-only contract calls on the configured EVM network.\n3. For a regular feed, it calls:\n   - `decimals()` to learn the feed’s precision.\n   - `latestAnswer()` to obtain the current integer answer.\n4. The workflow applies the decimals to present the regular-feed answer as a scaled value, then returns the resulting feed entries as JSON.\n5. For an MVR feed, it calls:\n   - `latestBundle()` to retrieve the encoded bundle.\n   - `bundleDecimals()` to obtain precision metadata for bundle fields.\n6. It decodes the bundle bytes according to the feed’s field schema, scales numerical values, and returns the decoded fields as a structured JSON result.\n\nIn short: select **regular** templates for one price/value per feed contract; select **MVR** templates when the feed is intentionally published as a typed, multi-field bundle.",
      "reason": "The table includes both languages and all four exact regular and MVR template IDs."
    },
    {
      "outcome": "The response explains that the workflow runs every 10 minutes on a cron trigger, uses the CRE chain reader to call the feed contract, and returns scaled regular-feed values or decoded MVR fields.",
      "pass": true,
      "evidence": "Choose based on the shape of data you need:\n\n- **Regular Data Feeds**: use these for a conventional **single-value feed**, such as BTC/USD or ETH/USD. The workflow reads `decimals()` and `latestAnswer()` from each feed contract, scales the answer using its decimals, logs it, and returns a JSON array of feed results.\n- **MVR (Multiple Variable Response) Data Feeds**: use these when one feed supplies a **bundle of multiple related values**. The workflow reads `latestBundle()` and `bundleDecimals()` from a `BundleAggregatorProxy`, decodes the returned bytes into typed fields, scales numeric fields as appropriate, and returns a structured JSON object. The provided MVR example uses the S&P Global SSA EURC feed on Base.\n\n## Template choices\n\n| Use case | Language | CRE CLI template ID |\n|---|---|---|\n| Regular single-value feeds | TypeScript | `read-data-feeds-ts` |\n| Regular single-value feeds | Go | `read-data-feeds-go` |\n| MVR bundled multi-value feeds | TypeScript | `read-mvr-data-feeds-ts` |\n| MVR bundled multi-value feeds | Go | `read-mvr-data-feeds-go` |\n\n## How the cron-based chain-reader flow works\n\n1. A **cron trigger fires every 10 minutes**.\n2. The CRE workflow uses the **CRE chain reader** to make read-only contract calls on the configured EVM network.\n3. For a regular feed, it calls:\n   - `decimals()` to learn the feed’s precision.\n   - `latestAnswer()` to obtain the current integer answer.\n4. The workflow applies the decimals to present the regular-feed answer as a scaled value, then returns the resulting feed entries as JSON.\n5. For an MVR feed, it calls:\n   - `latestBundle()` to retrieve the encoded bundle.\n   - `bundleDecimals()` to obtain precision metadata for bundle fields.\n6. It decodes the bundle bytes according to the feed’s field schema, scales numerical values, and returns the decoded fields as a structured JSON result.\n\nIn short: select **regular** templates for one price/value per feed contract; select **MVR** templates when the feed is intentionally published as a typed, multi-field bundle.",
      "reason": "The workflow description includes the 10-minute cron, CRE chain-reader contract calls, regular scaling, and MVR decoding."
    }
  ],
  "overallPass": true,
  "providerCalls": 2
}