sdd-tasks

sdd-tasks

Breaks down the design into an atomic, ordered, and verifiable task plan.

Triggers: /sdd-tasks <change-name>, task plan, break down implementation, task breakdown, sdd tasks

Skill resolution

Project-local → global, in order:

1. .claude/skills/sdd-tasks/SKILL.md     (project-local — highest priority)
2. ~/.claude/skills/sdd-tasks/SKILL.md   (global catalog — fallback)

See docs/SKILL-RESOLUTION.md for the full algorithm.

These notes bind the tool-neutral logic above to the Claude Code harness. They are the EXECUTION layer that does not port to other tools.

Step 0 preload — follow skills/_shared/sdd-phase-common.md Section F (Project Context Load) and Section G (Spec Context Preload). Both are non-blocking.
Artifact reads (Step 1) — use the harness Read("docs/sdd/{change-name}/<artifact>.md") tool for design, spec, and proposal. If a disk file is absent (legacy change), fall back to mem_search(query: "sdd/{change-name}/<artifact>") → mem_get_observation(id). If neither exists and Engram is not reachable, the content is passed inline from the orchestrator.
Task-plan persistence (Step 5) — disk-primary, two steps, disk FIRST: Write("docs/sdd/{change-name}/tasks.md", <full tasks markdown>), then mem_save a pointer + summary with topic_key: sdd/{change-name}/tasks, type: architecture, project: {project}, content = "path: docs/sdd/{change-name}/tasks.md\n\n{1-3 sentence summary}". If Engram MCP is not reachable: still write the disk artifact (the source of truth); skip the engram pointer only. The Design: / spec-review references inside the generated tasks.md use the engram:sdd/[name]/... pointer form.
Orchestrator output — return the neutral Output Contract as the JSON block the orchestrator expects; the artifacts entry is engram:sdd/{change-name}/tasks and the manual next step is /sdd-apply. Do not format the next step yourself — the orchestrator renders it with the slash prefix.