airmosaic-ai

GCAM Skill Data Boundary

Belongs In This Skill

• XQuery/XPath templates and patterns
• Python/R scripts that query BaseX via parameterized calls
• Reference documentation about GCAM output structure, configuration, and input chain
• Flat parameter JSON for registered run-tool adapters, and clearly marked non-executable manual plans
• Unit and scenario naming conventions

Does NOT Belong In This Skill

• GCAM model releases, executables, or Java binaries
• BaseX database directories or their contents
• Scenario output files (CSV, XML, PNG): keep in ${AIRMOSAIC_ASSETS_DIR}/results/<project_id>/<run_id>/<stage>/<tool_id>/
• Raw input CSV data used to build model scenarios
• Generated intermediate files (sector registries, method registries)
• Project-specific analysis outputs
• Large data archives (zip, tar, parquet)
• Personal browser profiles, cookies, or session tokens
• Private download URLs, API keys, or credentials
• Absolute local paths containing user home or workspace names

Where Outputs Go

Enabled adapters allocate new files under:

${AIRMOSAIC_ASSETS_DIR}/results/<project_id>/<run_id>/<stage>/<tool_id>/
${AIRMOSAIC_ASSETS_DIR}/results/<project_id>/<run_id>/records/<tool_id>/record.json

The two registered GCAM R tools use run-tool task outputs and sanitized receipts; plotting receives a runner-owned UNIT_PRICE_OUT_DIR. Direct standalone calls bypass these protections. Keep all artifacts local; only reviewed summaries may be shared. No automatic cache promotion or file cleanup occurs.

Where GCAM Release Package Lives

The GCAM release package (containing exe/, input/, output/, libs/) must be installed separately by the user. Use:

$env:GCAM_RELEASE_DIR = "<path to release package>"

Store the private mapping in ${AIRMOSAIC_CONFIG_DIR} and load it explicitly. Authentication state, if separately needed, belongs under ${AIRMOSAIC_AUTH_DIR} or in a credential manager. This skill does not provide a credential loader.

This site is open source. Improve this page.