tecton

name	tecton
description	Run Tecton plan and tests via Pants in the data-science repo. Handles long-running commands with proper output capture to avoid truncation.

Tecton Operations Skill

Critical: Output Handling for Long Commands

Tecton commands (especially tecton-plan) produce very long output that gets truncated. Capture to a literal file under /tmp/opencode/ and inspect it with the Read tool / rg:

PANTS_CONCURRENT=True pants tecton-plan --workspace=<workspace> --skip-tests 2>&1 | tee /tmp/opencode/tecton-plan.txt; echo "EXIT_CODE: ${PIPESTATUS[0]}"

Then inspect /tmp/opencode/tecton-plan.txt:

• Use the Read tool (offset/limit) to view sections.
• Use rg "error|Error|ERROR" /tmp/opencode/tecton-plan.txt to find failures.

Key points:

• Write the capture file to a literal /tmp/opencode/... path. tee to a $(mktemp) variable is blocked by bash-guard (unresolved-variable write target).
• Capture the real exit code with echo "EXIT_CODE: ${PIPESTATUS[0]}" — $? would report tee's exit code, not pants'.
• Inspect with the Read tool or rg; no manual cleanup needed (/tmp/opencode is ephemeral, and rm would trigger a permission prompt).

Environment

• This skill prefixes pants invocations with PANTS_CONCURRENT=True so multiple plan/test runs can proceed in parallel. The PANTS_CONCURRENT=True pants tecton-plan * and PANTS_CONCURRENT=True pants test * forms are allow-listed in opencode.jsonc, so they run without a permission prompt.
• All commands must be run from the data-science repo root.

Workspaces for Testing

• checkoutandsale-v1-staging: Staging workspace - check against this first
• bence-dev: Personal dev workspace for clarifying issues if staging fails with confusing errors

If staging passes, no need to check against bence-dev. Use bence-dev only when staging errors are unclear.

Operations

tecton-plan

Runs tecton plan against a workspace. Use --skip-tests to speed up iteration.

# Plan against staging (check this first)
PANTS_CONCURRENT=True pants tecton-plan --workspace=checkoutandsale-v1-staging --skip-tests 2>&1 | tee /tmp/opencode/tecton-plan-staging.txt; echo "EXIT_CODE: ${PIPESTATUS[0]}"

# Plan against dev workspace (only if staging errors are unclear)
PANTS_CONCURRENT=True pants tecton-plan --workspace=bence-dev --skip-tests 2>&1 | tee /tmp/opencode/tecton-plan-dev.txt; echo "EXIT_CODE: ${PIPESTATUS[0]}"

Inspect the capture files with the Read tool or rg.

tecton-test-run

Run Tecton tests using the pants test goal with --tecton-test-run flag.

# Run all tecton tests in a directory
PANTS_CONCURRENT=True pants test --tecton-test-run projects/features/repos/sig_tecton/tests/sig_tecton/datasources/orders/::

# Run a specific test file
PANTS_CONCURRENT=True pants test --tecton-test-run projects/features/repos/sig_tecton/tests/sig_tecton/datasources/orders/test_orders_datasources.py

# Run a specific test function
PANTS_CONCURRENT=True pants test --tecton-test-run projects/features/repos/sig_tecton/tests/path/to/test_file.py -- -k test_function_name

For long test runs, capture and inspect:

PANTS_CONCURRENT=True pants test --tecton-test-run projects/features/repos/sig_tecton/tests/:: 2>&1 | tee /tmp/opencode/tecton-test.txt; echo "EXIT_CODE: ${PIPESTATUS[0]}"