LCA Skills catalog

Reference map for the 28 canonical LCA Skills and the workflows they support.

Orchestration and execution

Skill	Use it for
lca-expert	End-to-end routing and uncertain multi-domain requests
lca-intake	Decision context, risk/claim classification, and missing inputs
lca-setup	Study workspace initialization
lca-plan	Reviewable task plans and acceptance criteria
lca-work	Execute approved study tasks
lca-autopilot	Explicitly authorized bounded end-to-end execution
lca-calculate	Controlled calculation after pre-calculation gates
lca-debug	Reproduce and diagnose broken or implausible models/results
lca-validate	Gate and readiness checks
lca-simplify	Reduce complexity while preserving decision-relevant behavior
lca-handoff	Pause, resume, and transfer durable state
lca-doctor	Diagnose package, host, Python, MCP, and optional tools
lca-compound	Turn validated lessons into reusable references and evals

Methodology, data, review, and reporting

Skill	Use it for
lca-scope	Goal/scope, function, boundary, modeling rules, and review route
lca-inventory	Foreground process and inventory work
lca-data	Dataset selection, mapping, provenance, and licensing
lca-impact	LCIA selection, mapping, characterization, normalization/weighting
lca-interpret	Hotspots, scenarios, sensitivity, uncertainty, and conclusions
lca-review	QA, red-team review, and release gating
lca-report	Reports, claims register, and release package
lca-research	Resolve current standards, methods, data, or evidence gaps

Tools, study types, and domains

Skill	Use it for
lca-openlca	openLCA setup, calculation, import/export, and diagnostics
lca-brightway	Brightway modeling, matrices, scenarios, and uncertainty
lca-greet	A named GREET product, release, and pathway
lca-prospective	Future/emerging technology and temporal/spatial scenarios
lca-epd-pcf	PCF, EPD, PCR/PEFCR, and PEF-style work
lca-sector	Domain/process-engineering checks
lca-organization-social	O-LCA, OEF, Scope 3 bridges, portfolio/social LCA

Invocation depends on the host

Claude Code plugin mode uses namespaced slash commands, while standalone Claude Code uses unnamespaced commands. ChatGPT and Claude.ai normally activate the consolidated lca-expert upload automatically. Codex invocation must be confirmed in the installed host/version.

Warning

Do not copy Claude Code slash-command syntax into ChatGPT, Claude.ai, or Codex instructions unless that host actually presents the same syntax.