LEED Tracker

Overview

Build a LEED v4.1 BD+C: New Construction scorecard for a project. The user declares pursuit status for each credit; this skill computes the running point total and the predicted certification level. Output is a CSV (one row per credit) plus a markdown summary in docs/leed/. This replaces the manual spreadsheet pass at every project's pre-design phase, kept evergreen through CD.

This skill does not decide whether a credit is achievable for a project — that is a strategy call by the LEED-AP and the design team. The skill organizes the structure, performs the arithmetic, cites the point matrix verbatim from references/, and flags which credits the user has not yet committed on.

When to use

Trigger on:

• "Make a LEED scorecard for this project."
• "Score this project for LEED v4.1."
• "Update the LEED tracker — we just confirmed EA Cr 2 is in."
• Pasted sustainability targets ("project targets LEED Silver — track progress").

Do NOT trigger on:

• "What credits can my project earn?" — that's a strategy question, not a tracker request.
• LEED v4 (older) or v5 (newer) — out of scope.
• Other rating systems (WELL, Living Building Challenge, Passive House) — those have separate trackers in the catalog.

Workflow

Step 1 — Bootstrap if needed

Check whether docs/leed/index.md exists. If not, run:

python scripts/init_workspace.py

The script creates docs/leed/ and an empty tracker log. No external Python dependencies. Confirm with the user before bootstrapping if docs/ does not already exist.

Step 2 — Confirm rating system

State explicitly to the user:

"This skill scores against LEED v4.1 BD+C: New Construction (110 points; Cert 40 / Silver 50 / Gold 60 / Platinum 80). If your project is on a different rating system or version, this skill is the wrong tool. Continue?"

Do not proceed silently if the user has not confirmed the rating system.

Step 3 — Gather project context

Read docs/project.md if present. Otherwise ask the user for:

• project_name
• target_certification_level ("Certified" / "Silver" / "Gold" / "Platinum" / "TBD")
• phase ("pre-design" / "SD" / "DD" / "CD" / "construction" / "operation")

Step 4 — Initialize a fresh scorecard, or update an existing one

If a scorecard CSV exists at docs/leed/<project-slug>-leed-scorecard.csv, load it and present current totals before editing. Otherwise initialize from assets/leed-v41-bdc-nc-scorecard-template.csv, which is pre-seeded with every credit's name, category, max-points, and prereq flag — sourced verbatim from references/leed_v41_bdc_nc_points.md.

Step 5 — Walk the credit list with the user

For each credit, the user declares one of:

• Yes — committed to pursuit; counts toward score.
• Maybe — under consideration; counted only in the optimistic total.
• No — not pursuing; zero contribution.
• N/A — not applicable to this project type.

For credits with point options (e.g. EA Credit 2 has multiple paths each yielding 1–18 points), the user states the points-attempted; the skill verifies that against the max published in references/.

For prerequisites (P), they must be Yes or N/A. If marked N/A, the user must justify; if No, the project is non-compliant — flag.

Step 6 — Compute scores

Run:

python scripts/compute_score.py \
  --scorecard docs/leed/<project-slug>-leed-scorecard.csv \
  --out docs/leed/.cache/<project-slug>-score.json

The script computes:

• Committed points (Yes only).
• Optimistic points (Yes + Maybe).
• Predicted certification level for both totals using the v4.1 BD+C thresholds (40 / 50 / 60 / 80).
• Per-category subtotals.
• Prerequisite status (all met / failures listed).

The script reads thresholds and max points from references/leed_v41_bdc_nc_points.md — never hard-coded into the script.

Step 7 — Write the markdown summary

Beside the CSV, write docs/leed/<project-slug>-leed-summary.md:

• Header with project name + target + phase + last-updated date.
• Committed total, optimistic total, predicted cert level for each.
• Per-category subtotal table.
• List of credits marked Maybe (next-decision queue).
• Prereq compliance status.
• Citation footer: standard version (LEED v4.1 BD+C: NC), reference file path, USGBC Reference Guide pointer.

Step 8 — Update the project LEED index

Run:

python scripts/append_to_index.py \
  --date 2026-05-08 \
  --project "<project-slug>" \
  --target "<Silver|Gold|Platinum|Certified>" \
  --phase "<phase>" \
  --committed <committed-points> \
  --optimistic <optimistic-points> \
  --cert-predicted "<level>" \
  --csv "<project-slug>-leed-scorecard.csv" \
  --summary "<project-slug>-leed-summary.md"

Appends a row to docs/leed/index.md. Do NOT edit index.md with Edit.

Step 9 — Report

Tell the user:

• Path to scorecard CSV + summary.
• Committed / Optimistic point totals + predicted cert level for each.
• Top 5 credits marked Maybe (highest-leverage decisions).
• Any prereq concerns.
• Reminder: this is structural tracking, not eligibility judgment — verify each Yes against USGBC LEED v4.1 BD+C Reference Guide credit requirements.

Rules

• Every row in the scorecard CSV must reference a credit that exists in references/leed_v41_bdc_nc_points.md. No invented credits.
• Max points per credit are quoted from the references file; the skill never increases a credit's max (e.g. EA Credit 2 max = 18 — do not let the user enter 20).
• Certification thresholds come from references/leed_v41_bdc_nc_points.md, never from training data.
• Score arithmetic runs through compute_score.py, never inline math by Claude.
• Output files live in docs/leed/ of the user's project.
• Standard version (LEED v4.1 BD+C: New Construction) and last-verified date are reported in every summary.

Anti-patterns

• Adding a credit not in references/. If the user asks about a credit by a name not in the references, treat as user typo and propose the closest match — do not invent.
• Letting the user record more points than the credit's max.
• Predicting a higher cert level than the thresholds support.
• Recommending which credits to pursue. This skill is structural — strategic credit selection belongs to the LEED-AP.
• Editing docs/leed/index.md directly with Edit.
• Treating LEED v4 / v5 / different rating systems as scoreable here.

Resources

• scripts/init_workspace.py — bootstrap docs/leed/
• scripts/compute_score.py — sum points, evaluate cert level, validate against references
• scripts/append_to_index.py — append a row to docs/leed/index.md
• references/_provenance.md — source, version, last-verified
• references/leed_v41_bdc_nc_points.md — full credit list with point values, prereqs, thresholds
• references/leed_categories.md — category structure, point allocation per category
• assets/leed-v41-bdc-nc-scorecard-template.csv — pre-seeded scorecard (one row per credit)
• assets/leed-index-template.md — header table for docs/leed/index.md