Action Playbook

Use this page to turn sustainability principles into everyday engineering and content decisions. Start small, measure consistently, and improve every cycle.

Back to Home View Source

What this page is for

This playbook translates sustainability principles into repeatable actions for humans and AI coding agents. It is intentionally practical: short checklists, clear ownership, and measurable outcomes.

Operating principles

  • Measure before and after changes.
  • Prefer simpler, lower-compute solutions.
  • Keep critical paths lean; move heavy work out of request paths.
  • Treat sustainability as an operational quality, like reliability and security.
  • Improve continuously rather than waiting for perfect data.

Rollout approach

Phase 1: Starter commitment + CI gates first

  • Publish a short public commitment in SUSTAINABILITY.md.
  • Enforce CI gates for accessibility and sustainability budgets on every PR.
  • Run Lighthouse on every PR against the live site URL.
  • Start with realistic thresholds (for example 90+) to stabilize signal quality.

Phase 2: Full template adoption after ownership is clear

  • Add full sections for ownership, scope, exceptions, and governance.
  • Establish metric baselines and monthly targets per owner.
  • Tighten CI and Lighthouse gates progressively toward 100/100/100/100.

Weekly workflow (team)

1) Plan with impact in mind

  • Tag work items with sustainability impact labels (for example: performance-budget, third-party-impact, ai-usage, grid-aware).
  • Identify one expected effect for each change (for example: fewer requests, smaller page weight, reduced CI minutes).
  • Assign an owner for measurement and follow-up.

2) Build with low-footprint defaults

  • Prefer static generation and caching where possible.
  • Remove unused scripts, CSS, fonts, and images before adding new assets.
  • Use efficient media formats and right-sized images.
  • Keep third-party dependencies to a justified minimum.

3) Review with a sustainability gate

  • Verify no budget regressions in CI for page weight, request count, or build overhead.
  • Require explicit justification for new third-party scripts and AI-heavy workflows.
  • Block merges when thresholds regress beyond agreed limits.
  • Review Lighthouse trendlines each month and ratchet thresholds upward.

Monthly cadence (team)

1) Baseline and trend review

  • Review key metrics trend lines (not only single-point snapshots).
  • Compare current values against targets and previous month.
  • Document top regressions and top improvements.

2) Focused optimization sprint

  • Pick the highest-impact bottleneck first (largest transfer, noisiest script, heaviest job).
  • Ship one or two concrete fixes with before/after evidence.
  • Update standards/checklists to prevent recurrence.

3) Policy and knowledge updates

  • Update this policy and related docs as architecture and tools change.
  • Share wins, misses, and trade-offs with the broader team.
  • Keep training lightweight but recurring.

AI agent playbook

Default behavior

  • Prefer deterministic scripts and templates over model calls.
  • Use the smallest suitable model and shortest useful context.
  • Reuse cached outputs and avoid duplicate prompts.

Allowed high-value AI use

  • Migration planning and scoped refactoring analysis.
  • Summarization where no deterministic equivalent is available.
  • Triage tasks that reduce repeated human effort.

Restricted use

  • No routine always-on AI generation in CI.
  • No large-context repetition for trivial formatting work.
  • No AI call when a local deterministic tool can produce the same result.

AI logging and accountability

  • Record AI-assisted tasks in PR descriptions when relevant.
  • Track approximate AI call volume per PR or issue.
  • Review usage trends monthly and target reductions.

Carbon-aware operations

CO2.js measurement guidance

Use CO2.js to keep carbon estimation consistent across teams and PRs.

  • Use perByte(bytes, green) when you want emissions for a single transfer event (for example: an upload, API payload, or one asset request).
  • Use perVisit(bytes, green) when you want an estimated emissions figure for one full page view based on total transfer bytes.
  • If you need explainability for audits, use perByteTrace or perVisitTrace to capture the variables used in the estimate.

Practical decision pattern

  • Calculate transfer estimates with CO2.js from measured bytes.
  • Fetch current/local grid signal via a grid-aware source.
  • If grid intensity is higher than average, serve a lighter variant (fewer heavy assets, defer non-critical media).
  • If grid intensity is lower than average, allow richer non-critical enhancements.

Reference implementations and docs:

  • https://developers.thegreenwebfoundation.org/co2js/overview/
  • https://observablehq.com/@greenweb/co2-js-playground
  • https://gaw.greenweb.org/grid-intensity/
  • https://github.com/thegreenwebfoundation/grid-intensity
  • https://www.thegreenwebfoundation.org/news/a-new-api-for-grid-aware-websites-and-beyond/

Time shift

  • Schedule deferrable jobs in lower-carbon windows when practical.
  • Define safe delay bounds to protect delivery commitments.

Space shift

  • Prefer lower-carbon hosting regions where architecture allows.
  • Use self-hosted runners for region-aware background jobs when needed.

Keep constraints explicit

  • Document platform constraints clearly (for example, limited control over hosted runner location).
  • Focus optimization efforts where control is real and measurable.

Suggested scorecard

Track at least these indicators:

  • Core template page weight.
  • Third-party request count.
  • Build and CI compute minutes.
  • AI calls per PR (or per release).
  • Number of deferrable jobs shifted in time or space.

30-day starter checklist

  • Week 1: Define 3-5 metrics and establish baseline.
  • Week 2: Add CI checks for budgets and regressions.
  • Week 3: Remove one major source of avoidable transfer or compute.
  • Week 4: Publish results and set next-month targets.

Progress over perfection: repeat the cycle and make each month lighter than the last.