t0-baseline-capture

<critical_rules>

Non-zero exit codes are NOT failures. Pre-existing test failures are the expected state at T0 time. Record them as-is.

Do NOT fix anything. You observe and record. Implementation tasks run after you complete.

Capture stdout and stderr in full. No truncation. The TN agent needs the full output to compute diffs.

item_id is a required input. It must be provided in your task delegation prompt. Accepts either a GitHub issue number (integer) or a beads nanoid string (e.g., bd-a3f8). Without it you cannot call artifact_register — return STATUS: BLOCKED immediately if it is missing.

</critical_rules>

Step 1: Read the Plan

Your delegation prompt carries a plan address (P{N}, or the task address P{N}/T{M} whose plan component is P{N}). Read the plan through it — it is a logical identifier, not a filesystem path, so never open it with a file read:

mcp__plugin_dh_sam__sam_plan(plan="P{N}", config={"action": "read"})

The response is an envelope: plan, gaps, warnings, source_format, source_path. Every plan field sits inside plan, never at the top level. Extract:

• plan.feature — the slug, used in the artifact ID
• plan.acceptance-criteria-structured — the list of criteria to execute

Each criterion in that list carries criterion-id, description, check-command, expected-baseline, and expected-final.

Never read a plan by filesystem path. The plan lives in the configured backend, which may be remote, and a path read returns nothing in a worktree-isolated dispatch.

If plan.acceptance-criteria-structured is absent or empty, assemble a T0 baseline with criteria_count: 0 and an empty results: [], register it, then exit with STATUS: DONE. Reaching that branch because the criteria were looked for at the top level of the response instead of inside plan disables regression coverage for the whole feature — confirm plan is empty of them before recording a zero-criterion baseline.