@raishin/vanguard-frontier-agentic 1.8.0 → 2.0.0

package/skills/finops/carbon-cost-pair/README.md

@@ -0,0 +1,25 @@
+# Carbon Cost Pair
+
+A FinOps skill that pairs cloud spend values with kgCO2e carbon estimates by region and service category, with mandatory confidence labels and source citations for CSRD and SEC climate disclosure.
+
+## Purpose
+
+For every cloud spend line item, produce a paired kgCO2e estimate using the best available carbon factor for the cloud region and service category. Confidence is labeled as `vendor-published`, `third-party`, or `estimated`, and every output includes a source URL and data period acknowledgment.
+
+## Allowed tools
+
+`Read` `Grep` `Glob` `WebFetch`
+
+## Usage
+
+**Single line item:** Provide cloud provider, region, service category, and dollar value (e.g., "Pair $1,200 of AWS us-east-1 compute spend with a carbon estimate"). The skill fetches the latest available carbon factor, applies it, and returns a labeled kgCO2e output with FOCUS tag mapping.
+
+**Multi-line comparison:** Provide multiple regions or providers (e.g., "Compare carbon intensity for the same $10,000 compute spend across AWS us-east-1, Azure eastus, and GCP us-central1"). The skill returns one estimate block per line item plus a summary comparison table.
+
+## Trust posture
+
+Read-only. No cloud credentials, billing account IDs, or tenant data accepted. Carbon factors are fetched from public provider sustainability pages or third-party sources. Provider data typically lags 60-90 days; the data period is stated in every output.
+
+Defaults to Scope 2 electricity-only, market-based factors. Location-based and Scope 1+2 available on request.
+
+See [SKILL.md](SKILL.md) for the full operating rules, confidence label decision tree, regulatory context, and output shape.

package/skills/finops/carbon-cost-pair/SKILL.md

@@ -0,0 +1,96 @@
+---
+name: carbon-cost-pair
+description: Pair every cloud spend value with a kgCO2e estimate by region and service category for CSRD and SEC climate disclosure. Input is region, service category, and dollar value; output is paired kgCO2e with confidence label (vendor-published, third-party, or estimated) and source citation. Defaults to electricity-only Scope 2 market-based factors.
+allowed-tools: Read Grep Glob WebFetch
+metadata:
+  author: "github: Raishin"
+  version: "0.1.1"
+  updated: "2026-05-13"
+  category: finops
+  lifecycle: experimental
+---
+
+# Carbon Cost Pair
+
+## Purpose
+
+For every cloud spend line item, produce a paired kgCO2e estimate suitable for CSRD and SEC climate disclosure. The skill pairs a dollar value (or resource quantity) with the best available carbon emissions factor for the given cloud region and service category, and returns the estimate with a mandatory confidence label and source citation.
+
+No cloud credentials accepted. Carbon factors are fetched from provider sustainability pages or third-party sources.
+
+## When to use
+
+Use this skill when:
+
+- The user needs to pair a cloud spend or usage figure with a carbon estimate for sustainability reporting (CSRD Article 29a, SEC climate rule, GHG Protocol Scope 2)
+- The user wants to compare the carbon intensity of the same workload across cloud regions or providers
+- The user wants to understand the confidence level of a carbon estimate before including it in a regulatory disclosure
+
+## Operating rules
+
+- **Default to electricity-only Scope 2, market-based.** Unless the caller specifies otherwise, compute Scope 2 emissions using market-based electricity factors (renewable energy certificate or PPA-adjusted). Location-based factors are available on request.
+- **Fetch live data first.** Use WebFetch to retrieve the latest carbon factor data from the provider's sustainability API or dashboard before falling back to documentation-based estimates.
+- **Confidence labels mandatory.** Every kgCO2e output must carry exactly one label:
+  - `vendor-published` — factor sourced from the provider's own sustainability API, dashboard, or data export within this session (include source URL and ISO 8601 timestamp)
+  - `third-party` — factor sourced from Electricity Maps, Cloud Carbon Footprint project, or equivalent third-party source (include source URL and timestamp)
+  - `estimated` — factor derived from regional grid average or published industry emission factors where no provider or third-party data is available (state the factor, source, and assumptions)
+- **No credentials accepted.** No cloud credentials, billing account IDs, tenant data, or sustainability API tokens are accepted or required.
+- **Acknowledge data lag.** Provider carbon data typically lags 60-90 days or more. State the data period whenever citing vendor-published factors.
+- **FOCUS column mapping.** There are no native FOCUS v1.2 carbon columns. Represent carbon estimates as FOCUS tags: `Tags/kgco2e`, `Tags/carbon_confidence`, `Tags/carbon_source`.
+- **Load references only when needed.**
+
+## Input
+
+| Field | Required | Description |
+|---|---|---|
+| Cloud provider | Yes | AWS, Azure, Google Cloud, OCI, or multi-cloud |
+| Region | Yes | Provider region name (e.g., `us-east-1`, `eastus`, `us-central1`, `ap-sydney-1`) |
+| Service category | Yes | Compute, Storage, Networking, AI and Machine Learning, Databases, or Other |
+| Dollar value ($) | Yes | The spend amount to pair with a carbon estimate |
+| Resource quantity | Optional | If available, the underlying resource quantity (e.g., vCPU-hours, GiB-hours, GB transferred) for a more precise estimate |
+| Scope preference | Optional | Default: Scope 2 market-based. Options: `scope2-market`, `scope2-location`, `scope1+2` |
+
+## Output shape
+
+Return, per line item:
+
+```
+Region:           <provider region>
+Service category: <category>
+Dollar value:     $<amount>
+kgCO2e estimate:  <value> kgCO2e  [<confidence label>]
+Scope:            Scope 2 (market-based)  |  Scope 2 (location-based)
+Data period:      <e.g., Q3 2025 or "Current as of 2026-05-13">
+Source:           <URL>  (fetched: <ISO 8601 timestamp>  |  documentation date: <date>)
+Methodology note: <one sentence describing the emission factor used>
+
+FOCUS tags:
+  Tags/kgco2e:           <value>
+  Tags/carbon_confidence: <vendor-published | third-party | estimated>
+  Tags/carbon_source:     <URL>
+```
+
+If multiple regions or service categories are provided, return one block per line item followed by a summary table.
+
+## Confidence label decision tree
+
+1. Does the cloud provider publish a region-level carbon factor or API for this region and service?
+   - Yes, and it was fetched this session → `vendor-published`
+   - Yes, but the fetch failed or data is not current → proceed to step 2
+2. Is Electricity Maps data available for the grid region, or is Cloud Carbon Footprint data available for this provider/region?
+   - Yes → `third-party`
+3. Fall back to IEA regional grid average emission factor or the provider's country-level average.
+   - → `estimated`; state the factor used (e.g., `0.386 kgCO2e/kWh — IEA 2023 US average`)
+
+## Regulatory context
+
+- **CSRD (Corporate Sustainability Reporting Directive)**: requires disclosure of Scope 1, 2, and 3 emissions under ESRS E1. Cloud compute is typically Scope 2 (purchased electricity) and may contribute to Scope 3 Category 1 (purchased goods/services) for customers.
+- **SEC Climate Rule**: requires material climate-related disclosures including Scope 1 and 2 emissions for large accelerated filers.
+- **GHG Protocol**: defines market-based and location-based approaches for Scope 2. Both may be required; market-based is the default here.
+
+## References
+
+Load these only when needed:
+
+- [Carbon data sources](references/sources.md) — provider sustainability API URLs, third-party data sources, and fallback options.
+- [Methodology](references/methodology.md) — market-based vs location-based Scope 2, data lag, and uncertainty labeling.

package/skills/finops/carbon-cost-pair/metadata.json

@@ -0,0 +1,27 @@
+{
+  "id": "carbon-cost-pair",
+  "name": "Carbon Cost Pair",
+  "type": "skill",
+  "provider": "multi-cloud",
+  "harnesses": [
+    "codex",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Pair cloud spend values with kgCO2e estimates by region and service category for CSRD and SEC climate disclosure. Outputs confidence-labeled carbon estimates (vendor-published, third-party, or estimated) with source citations. Defaults to Scope 2 market-based electricity factors.",
+  "source_type": "original",
+  "official_docs": [
+    "https://aws.amazon.com/aws-cost-management/aws-customer-carbon-footprint-tool/",
+    "https://learn.microsoft.com/en-us/industry/sustainability/",
+    "https://cloud.google.com/carbon-footprint/docs"
+  ],
+  "security_notes": "No cloud credentials, billing account IDs, sustainability API tokens, or tenant-specific data are accepted or required. All carbon factors are fetched from public provider sustainability pages or third-party data sources. No write operations to any system are performed.",
+  "last_verified": "2026-05-13",
+  "path": "skills/finops/carbon-cost-pair",
+  "author": "github: Raishin",
+  "version": "0.1.1",
+  "lifecycle": "experimental"
+}

package/skills/finops/carbon-cost-pair/references/methodology.md

@@ -0,0 +1,78 @@
+# Carbon Emission Methodology
+
+## Scope 2 emissions: market-based vs location-based
+
+Scope 2 emissions are indirect greenhouse gas emissions from the generation of purchased electricity consumed by an organization. The GHG Protocol defines two methods for Scope 2 accounting:
+
+### Market-based method
+
+Uses emission factors from contractual instruments that convey attributes of the electricity consumed, such as:
+- Renewable Energy Certificates (RECs) or Guarantees of Origin (GOs)
+- Power Purchase Agreements (PPAs) with specific generators
+- Supplier-specific emission factors published by utilities
+
+Effect: when a cloud provider purchases RECs or PPAs covering 100% of their electricity use in a region, the market-based Scope 2 emission factor for that region approaches zero. This is why hyperscaler-reported Scope 2 figures are often significantly lower than location-based figures.
+
+Default in this skill: **market-based**, because it reflects the cloud provider's actual contractual commitments and is the value reported in vendor sustainability disclosures.
+
+### Location-based method
+
+Uses average emission factors for the electricity grid(s) where consumption occurs, regardless of contractual instruments. These factors are typically published by grid operators or national energy agencies (e.g., EPA eGRID for US, AIB residual mixes for Europe).
+
+The location-based figure represents what emissions would have been if the electricity came from the average mix of sources on the local grid.
+
+Both methods may be required for CSRD (ESRS E1) and GHG Protocol dual-reporting.
+
+## Data lag
+
+Provider carbon data typically lags the billing period by 60-90 days or more:
+
+- AWS Customer Carbon Footprint Tool: approximately 3 months lag.
+- Microsoft Emissions Impact Dashboard: approximately 2-3 months lag.
+- Google Cloud Carbon Footprint: approximately 2 months lag.
+- Third-party (Electricity Maps historical): 24-hour lag for the previous day; monthly averages available sooner.
+
+Implication for disclosure: carbon estimates produced for a quarter's cloud spend will be based on emission factors from the same or a prior quarter. Always state the data period when citing vendor-published factors.
+
+## How to label uncertainty
+
+Every kgCO2e output must carry one of three confidence labels with the stated basis:
+
+| Label | When to use | What to state |
+|---|---|---|
+| `vendor-published` | Provider's own sustainability API or dashboard data was retrieved | Source URL, data period (e.g., "Q3 2025"), fetch timestamp |
+| `third-party` | Electricity Maps or Cloud Carbon Footprint data was used | Source URL, date of data, which grid zone was used |
+| `estimated` | Regional or national grid average used as fallback | Emission factor (kgCO2eq/kWh), source (e.g., IEA 2023), year of data |
+
+Do not present a kgCO2e figure without one of these labels. If a value cannot be computed with any of the three approaches, state that a carbon estimate is not available for this region or service category and explain why.
+
+## Conversion from $/spend to kgCO2e
+
+When only a dollar spend value is provided (not resource quantity), a spend-based emission factor is required. These are less precise than activity-based factors (kgCO2e per kWh or per vCPU-hour) because price varies by region, service type, and over time.
+
+If the provider publishes a spend-normalized factor (kgCO2e/$) for the relevant service or region, use it and label as `vendor-published`.
+
+If no spend-normalized factor is available, prefer an activity-based route:
+1. Estimate the resource quantity from the spend (e.g., divide by on-demand price per vCPU-hour to get vCPU-hours).
+2. Multiply vCPU-hours by the energy use per vCPU-hour (typical: 0.0025 kWh for a modern data center instance).
+3. Multiply energy use by the regional emission factor (kgCO2eq/kWh).
+
+State each step and label the result `estimated` unless the emission factor was fetched from a vendor or third-party source.
+
+## Scope 1 emissions (note)
+
+Scope 1 emissions (direct on-site combustion) from cloud data centers — backup generators, on-site equipment — are typically immaterial compared to Scope 2 and are not separately published by cloud providers per customer. Unless the caller specifically requests Scope 1, this skill produces Scope 2 only.
+
+## Regulatory framing
+
+| Framework | Requirement relevant to cloud |
+|---|---|
+| CSRD / ESRS E1 | Gross Scope 1, 2, and 3 by category; both market-based and location-based Scope 2; alignment with GHG Protocol |
+| SEC Climate Rule | Material Scope 1 and 2; Scope 3 if material or targeted; large accelerated filers required to include |
+| GHG Protocol Corporate Standard | Both market-based and location-based Scope 2 recommended; Scope 3 Category 1 (purchased goods/services) may include cloud |
+
+Cloud emissions typically appear in:
+- Scope 2 (electricity) for compute, storage, and networking in data centers
+- Scope 3 Category 1 (purchased goods and services) for cloud spend treated as a supply chain input
+
+This skill computes Scope 2 (electricity) only unless the caller specifies otherwise.

package/skills/finops/carbon-cost-pair/references/sources.md

@@ -0,0 +1,103 @@
+# Carbon Data Sources
+
+## Provider sustainability data (vendor-published)
+
+### AWS — Customer Carbon Footprint Tool
+
+| Resource | URL |
+|---|---|
+| Tool overview | https://aws.amazon.com/aws-cost-management/aws-customer-carbon-footprint-tool/ |
+| Documentation | https://docs.aws.amazon.com/cur/latest/userguide/ccft-overview.html |
+| Methodology | https://docs.aws.amazon.com/cur/latest/userguide/ccft-methodology.html |
+
+AWS Customer Carbon Footprint Tool provides estimated carbon emissions for AWS usage in metric tons of CO2 equivalent (MTCO2e) broken down by service, region, and time period. Data is available through the AWS console and the AWS Cost Explorer API.
+
+Important notes:
+- Data typically lags approximately 3 months (90 days).
+- Figures reflect AWS's own renewable energy purchases (market-based Scope 2).
+- Location-based figures are also available in the methodology documentation.
+- The tool does not expose a public unauthenticated API; for live fetch, use WebFetch against the public documentation and methodology pages, which publish the aggregate emission factors and regional renewable energy percentages.
+
+For unauthenticated access to AWS carbon methodology data, fetch from:
+- https://docs.aws.amazon.com/cur/latest/userguide/ccft-methodology.html
+- https://sustainability.aboutamazon.com/
+
+### Microsoft — Cloud for Sustainability Emissions API
+
+| Resource | URL |
+|---|---|
+| Industry sustainability hub | https://learn.microsoft.com/en-us/industry/sustainability/ |
+| Emissions Impact Dashboard | https://www.microsoft.com/en-us/sustainability/emissions-impact-dashboard |
+| Microsoft Sustainability Manager docs | https://learn.microsoft.com/en-us/industry/sustainability/sustainability-manager-overview |
+
+Microsoft publishes carbon data through the Emissions Impact Dashboard (Power BI report, not unauthenticated API) and the Microsoft Cloud for Sustainability APIs. For unauthenticated reference data, use:
+- https://www.microsoft.com/en-us/sustainability/reports (annual sustainability report with regional energy data)
+- https://learn.microsoft.com/en-us/azure/carbon-optimization/ (Azure carbon optimization documentation)
+
+Data lag: typically 2-3 months behind current billing.
+
+### Google Cloud — Carbon Footprint Export
+
+| Resource | URL |
+|---|---|
+| Documentation | https://cloud.google.com/carbon-footprint/docs |
+| Methodology | https://cloud.google.com/carbon-footprint/docs/methodology |
+| Region carbon data | https://cloud.google.com/sustainability/region-carbon |
+
+Google Cloud Carbon Footprint provides gross and net carbon emissions per product and region. The regional carbon data page publishes the carbon-free energy percentage per region, which can be combined with regional grid intensity to estimate net Scope 2 emissions.
+
+For unauthenticated reference:
+- https://cloud.google.com/sustainability/region-carbon — carbon-free energy percentages per region; fetch this page with WebFetch.
+- https://cloud.google.com/carbon-footprint/docs/methodology — describes the market-based and location-based calculation approach.
+
+### OCI — Sustainability Dashboard
+
+| Resource | URL |
+|---|---|
+| OCI Sustainability overview | https://www.oracle.com/sustainability/ |
+| Oracle sustainability report | https://www.oracle.com/corporate/citizenship/ |
+
+OCI does not publish a per-region unauthenticated carbon API equivalent to AWS CCFT or Google Cloud Carbon Footprint. For OCI carbon estimates, fall back to the Electricity Maps regional data for the grid serving the OCI data center, or use the Cloud Carbon Footprint project's OCI emission factors.
+
+## Third-party sources
+
+### Electricity Maps
+
+| Resource | URL |
+|---|---|
+| Global carbon intensity map | https://www.electricitymaps.com/ |
+| API documentation | https://static.electricitymaps.com/api/docs/index.html |
+| Open data | https://github.com/electricitymaps/electricitymaps-contrib |
+
+Electricity Maps publishes real-time and historical grid carbon intensity (gCO2eq/kWh) for electricity grids worldwide. This is suitable for location-based Scope 2 calculations. The public API requires an API key for live data; the open data repository on GitHub contains historical zone data accessible without authentication.
+
+For unauthenticated WebFetch, use the public zone data from the GitHub repository.
+
+### Cloud Carbon Footprint Project
+
+| Resource | URL |
+|---|---|
+| Project site | https://www.cloudcarbonfootprint.org/ |
+| Emission factors repository | https://github.com/cloud-carbon-footprint/cloud-carbon-footprint |
+| Coefficient data | https://github.com/cloud-carbon-footprint/cloud-carbon-footprint/blob/trunk/packages/gcp/src/domain/GcpFootprintEstimationConstants.ts |
+
+The Cloud Carbon Footprint project publishes open-source emission coefficients for AWS, Azure, and GCP including per-region grid intensity and average embodied carbon for common instance types. These are community-maintained, not vendor-published, and should be labeled `third-party`.
+
+## IEA and national grid averages (fallback)
+
+If neither vendor nor third-party data is available for a region:
+
+| Source | URL |
+|---|---|
+| IEA CO2 emissions from electricity data | https://www.iea.org/data-and-statistics/data-product/emissions-factors-2023 |
+| EPA eGRID (US) | https://www.epa.gov/egrid |
+
+Use the regional or national grid average in kgCO2eq/kWh and multiply by the estimated kWh consumed. Label as `estimated` and state the factor and its source year.
+
+## Source selection priority
+
+When multiple sources are available, use the highest-confidence source available:
+
+1. Vendor-published (provider sustainability API or dashboard documentation for the specific region)
+2. Third-party (Electricity Maps or Cloud Carbon Footprint for the specific grid region)
+3. Estimated (IEA or EPA national/regional average, clearly labeled)

package/skills/finops/fetch-foundation-model-pricing/README.md

@@ -0,0 +1,25 @@
+# Fetch Foundation Model Pricing
+
+A FinOps skill that retrieves live public pricing for foundation models across major AI and cloud providers, returning structured tables with mandatory provenance labels and source timestamps.
+
+## Purpose
+
+Fetch current per-token, per-image, and per-GPU-hour prices from Anthropic, OpenAI, Google (Vertex AI), AWS Bedrock, Azure OpenAI Service, and OCI Generative AI. Supports single-model lookups and side-by-side comparative tables.
+
+## Allowed tools
+
+`Read` `Grep` `Glob` `WebFetch`
+
+## Usage
+
+**Single-model lookup:** Provide a model name and deployment target (e.g., "What does Claude Sonnet 4.5 cost per million input tokens on Bedrock?"). The skill fetches the live price, labels it with source URL and ISO 8601 timestamp, and returns a single-row price table.
+
+**Comparative table:** Provide two or more models or providers and a task type (e.g., "Compare GPT-4o, Claude Sonnet, and Gemini Pro on input and output token cost"). The skill fetches each price independently and builds a labeled multi-row comparison with cheapest/most-expensive summary rows.
+
+## Trust posture
+
+Read-only. No cloud credentials, billing account IDs, or tenant data accepted. All pricing pages are public and unauthenticated. Every price value carries a provenance label (`live-price`, `documentation-based`, `assumed`, or `excluded`) with source URL and fetch timestamp.
+
+FOCUS v1.2 column mapping is included for any cost estimate produced (BilledCost, EffectiveCost, ServiceCategory, ChargeCategory, SkuId, SkuPriceId).
+
+See [SKILL.md](SKILL.md) for the full operating protocol, pricing dimensions, and response shape.

package/skills/finops/fetch-foundation-model-pricing/SKILL.md

@@ -0,0 +1,85 @@
+---
+name: fetch-foundation-model-pricing
+description: Fetch live per-token, per-image, and per-GPU-hour prices for foundation models across Anthropic, OpenAI, Google, AWS Bedrock, Azure OpenAI, OCI Generative AI, and Vertex AI. Supports single-model lookup and comparative multi-provider tables. Every price is labeled with source URL and ISO 8601 fetch timestamp. No credentials accepted.
+allowed-tools: Read Grep Glob WebFetch
+metadata:
+  author: "github: Raishin"
+  version: "0.1.1"
+  updated: "2026-05-13"
+  category: finops
+  lifecycle: experimental
+---
+
+# Fetch Foundation Model Pricing
+
+## Purpose
+
+Retrieve current public pricing for foundation models across the major AI/cloud providers and return structured, provenance-labeled output. Supports two modes:
+
+- **Single-model lookup**: fetch the current price for a specific model and deployment target (e.g., Claude Sonnet 4.5 on Anthropic direct, or on Bedrock).
+- **Comparative table**: build a side-by-side price comparison across two or more models or providers for the same task type (text, image, embedding, or GPU-hour).
+
+## When to use
+
+Use this skill when:
+
+- The user asks "how much does model X cost per token / per image / per GPU-hour"
+- The user wants to compare inference costs across two or more providers for the same model family or equivalent capability tier
+- The user needs to estimate monthly AI inference spend given a volume projection
+- The user wants to understand how context caching or batch pricing changes the effective cost curve
+- The user wants a FOCUS-aware cost breakdown (BilledCost, EffectiveCost) for AI inference line items
+
+## Operating rules
+
+- **Fetch live prices first.** Use WebFetch to retrieve prices from each provider's public pricing page before relying on any internal knowledge. AI model pricing changes frequently; stale numbers mislead.
+- **Label every price.** Each price value must carry exactly one provenance label:
+  - `live-price` — fetched from a provider's public pricing page or API within this session; include source URL and ISO 8601 timestamp.
+  - `documentation-based` — sourced from official documentation when a live fetch was not possible; note the documentation URL and its visible publication date.
+  - `assumed` — derived from an analogous model or tier when no direct published price exists; state the assumption explicitly.
+  - `excluded` — pricing that exists but was intentionally omitted from the output; state why.
+- **Include source URL and timestamp.** For every live-price value, state the exact URL fetched and the UTC timestamp of the fetch to the minute (e.g., `2026-05-13T14:32Z`).
+- **On-demand pricing only unless told otherwise.** Do not apply reserved capacity, committed use, or enterprise negotiated pricing unless the user explicitly requests it.
+- **No credentials required or accepted.** All provider pricing pages are public and unauthenticated. Never ask for API keys, billing account IDs, or tenant-specific data.
+- **FOCUS column mapping.** Where a cost estimate is produced, note the corresponding FOCUS v1.2 columns: `BilledCost` (what the provider charges), `EffectiveCost` (after credits or discounts), `ServiceCategory` (AI and Machine Learning), `ChargeCategory` (Usage), `SkuId` (model ID + deployment tier), `SkuPriceId` (price dimension: input-token / output-token / cached-token / image / gpu-hour).
+- **Load references only when needed.**
+
+## Pricing dimensions
+
+| Dimension | Unit | Applies to |
+|---|---|---|
+| Input tokens | per 1M tokens | All text models |
+| Output tokens | per 1M tokens | All text models |
+| Cached input tokens | per 1M tokens | Models with context caching (Anthropic, Gemini, Bedrock) |
+| Batch input tokens | per 1M tokens | Models with async batch mode (Anthropic, OpenAI, Bedrock) |
+| Batch output tokens | per 1M tokens | Models with async batch mode |
+| Images (input) | per image or per 1K images | Multimodal models |
+| GPU-hour | per hour per GPU type | Self-hosted / dedicated endpoints (Vertex AI, Bedrock provisioned throughput, Azure PTU) |
+
+## Providers in scope
+
+| Provider | Deployment target | Reference |
+|---|---|---|
+| Anthropic | Direct API | references/providers.md |
+| OpenAI | Direct API | references/providers.md |
+| Google | Vertex AI | references/providers.md |
+| AWS | Bedrock | references/providers.md |
+| Azure | Azure OpenAI Service | references/providers.md |
+| OCI | Generative AI Service | references/providers.md |
+
+## Response minimum
+
+Return, at minimum:
+
+- confirmed model name(s), provider(s), and deployment target(s)
+- pricing dimensions covered (input, output, cached, batch, image, GPU-hour)
+- line-item price table: model | provider | dimension | unit price | provenance label | source URL | fetch timestamp
+- for comparative tables: a summary row noting the cheapest and most expensive option per dimension
+- key assumptions (on-demand, no reserved/committed pricing, USD unless stated)
+- FOCUS column mapping for any cost estimate produced
+
+## References
+
+Load these only when needed:
+
+- [Provider pricing URLs](references/providers.md) — public pricing page URLs per provider for live WebFetch.
+- [Token economics](references/token-economics.md) — input/output/cached/batch pricing model, $/M token math, and context caching cost curve.

package/skills/finops/fetch-foundation-model-pricing/metadata.json

@@ -0,0 +1,30 @@
+{
+  "id": "fetch-foundation-model-pricing",
+  "name": "Fetch Foundation Model Pricing",
+  "type": "skill",
+  "provider": "multi-cloud",
+  "harnesses": [
+    "codex",
+    "claude-code",
+    "cursor",
+    "gemini",
+    "kiro",
+    "other"
+  ],
+  "summary": "Fetch live per-token, per-image, and per-GPU-hour prices for foundation models across Anthropic, OpenAI, Google, AWS Bedrock, Azure OpenAI, OCI Generative AI, and Vertex AI. Supports single-model lookup and comparative multi-provider tables with provenance labels.",
+  "source_type": "original",
+  "official_docs": [
+    "https://docs.anthropic.com/en/docs/about-claude/pricing",
+    "https://platform.openai.com/docs/pricing",
+    "https://aws.amazon.com/bedrock/pricing/",
+    "https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/",
+    "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+    "https://www.oracle.com/cloud/ai/generative-ai/"
+  ],
+  "security_notes": "All provider pricing pages are public and unauthenticated. Never accept or request API keys, billing account IDs, cost export access, tenant IDs, or any cloud credentials to fetch list prices.",
+  "last_verified": "2026-05-13",
+  "path": "skills/finops/fetch-foundation-model-pricing",
+  "author": "github: Raishin",
+  "version": "0.1.1",
+  "lifecycle": "experimental"
+}

package/skills/finops/fetch-foundation-model-pricing/references/providers.md

@@ -0,0 +1,88 @@
+# Provider Pricing URLs
+
+Use these URLs with WebFetch to retrieve live pricing. All pages are public and require no authentication.
+
+## Anthropic
+
+| Resource | URL |
+|---|---|
+| Pricing page | https://docs.anthropic.com/en/docs/about-claude/pricing |
+| Model overview | https://docs.anthropic.com/en/docs/about-claude/models/overview |
+
+Pricing dimensions available: input tokens ($/1M), output tokens ($/1M), prompt caching write ($/1M), prompt caching read ($/1M), batch input ($/1M), batch output ($/1M).
+
+Models published as of last verification: Claude Opus 4, Claude Sonnet 4.5, Claude Haiku 3.5, and prior generation models.
+
+## OpenAI
+
+| Resource | URL |
+|---|---|
+| Pricing page | https://platform.openai.com/docs/pricing |
+| Models list | https://platform.openai.com/docs/models |
+
+Pricing dimensions available: input tokens ($/1M), cached input tokens ($/1M), output tokens ($/1M). Batch API discounts published separately on the same page.
+
+Models published: GPT-4o, GPT-4o mini, o1, o3, o4-mini, text-embedding-3-large/small, DALL-E (per image).
+
+## AWS Bedrock
+
+| Resource | URL |
+|---|---|
+| Pricing page | https://aws.amazon.com/bedrock/pricing/ |
+| On-demand pricing table | https://aws.amazon.com/bedrock/pricing/#On-demand |
+| Batch inference | https://aws.amazon.com/bedrock/pricing/#Batch_inference |
+| Provisioned throughput | https://aws.amazon.com/bedrock/pricing/#Provisioned_throughput |
+
+Pricing dimensions available: on-demand input/output tokens ($/1000 tokens or $/1M tokens depending on model), batch input/output tokens, provisioned throughput model units ($/hour).
+
+Models published: Anthropic Claude family, Amazon Titan, Amazon Nova, Meta Llama, Mistral, Cohere, AI21 Jurassic, Stability AI.
+
+Note: Bedrock token prices are sometimes published per 1,000 tokens rather than per 1M. Normalize to $/1M for comparison by multiplying by 1,000.
+
+## Azure OpenAI Service
+
+| Resource | URL |
+|---|---|
+| Pricing page | https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/ |
+| Model catalog | https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models |
+
+Pricing dimensions available: pay-as-you-go input/output tokens ($/1K tokens), provisioned throughput units (PTU, $/hour/PTU). Prices vary by Azure region.
+
+Models published: GPT-4o, GPT-4o mini, o1, o3, o4-mini, text-embedding, DALL-E 3.
+
+Note: Azure OpenAI prices are often published per 1,000 tokens. Normalize to $/1M by multiplying by 1,000.
+
+## Google Vertex AI
+
+| Resource | URL |
+|---|---|
+| Generative AI pricing | https://cloud.google.com/vertex-ai/generative-ai/pricing |
+| Gemini model pricing | https://cloud.google.com/vertex-ai/generative-ai/pricing#gemini-models |
+| Embedding pricing | https://cloud.google.com/vertex-ai/generative-ai/pricing#embedding-models |
+
+Pricing dimensions available: input tokens ($/1M), output tokens ($/1M), context caching ($/1M per hour storage + discounted input tokens), grounding ($/1K queries), image input (per image or per 1K images for multimodal).
+
+Models published: Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 1.5 Pro/Flash, text-embedding-004, Imagen 3.
+
+## OCI Generative AI
+
+| Resource | URL |
+|---|---|
+| Service overview | https://www.oracle.com/cloud/ai/generative-ai/ |
+| Pricing page | https://www.oracle.com/cloud/ai/generative-ai/pricing/ |
+
+Pricing dimensions available: on-demand token pricing (input and output per 1M tokens), dedicated AI cluster pricing (unit/hour).
+
+Models published: Cohere Command R/R+, Meta Llama 3 family.
+
+Note: OCI Generative AI pricing may not be listed on the main service overview page. If the pricing page URL returns no results, also try https://www.oracle.com/cloud/price-list/ and search for "Generative AI".
+
+## Fetch strategy
+
+When using WebFetch against a pricing page:
+
+1. Fetch the canonical pricing URL listed above.
+2. Record the response timestamp in UTC (ISO 8601, e.g., `2026-05-13T14:32Z`).
+3. Extract the relevant model row(s) for the requested model name or family.
+4. If the page uses JavaScript to render a pricing table, it may return partial or placeholder content — in that case, label the result `documentation-based` and note the URL fetched.
+5. If the fetch fails entirely, fall back to the most recent documentation-based price known and label it `documentation-based` with the documentation URL.