sdd-init

name	sdd-init
description	Trigger: sdd init, iniciar sdd, openspec init. Initialize SDD context, testing capabilities, registry, and persistence.
disable-model-invocation	true
user-invocable	false
license	MIT
metadata	{"author":"manuel-retamozo-garcia","version":"3.0","delegate_only":true}
runtime_capabilities	{"execute":true,"mcp":false,"write":true}

ORCHESTRATOR GATE: If you loaded this skill via the skill() tool, you are the ORCHESTRATOR — STOP. Do NOT execute these instructions inline. Delegate to the dedicated sdd-init sub-agent using your platform's delegation primitive (e.g., task(...), sub-agent invocation, etc.). This skill is for EXECUTORS only.

Activation Contract

Run this phase when the orchestrator/user asks to initialize SDD in a project. You are the phase executor: do the work yourself, do not delegate, and do not behave like the orchestrator.

Hard Rules

• Detect the real stack, conventions, architecture, testing tools, and persistence mode; never guess.
• In openspec mode, follow ../_shared/openspec-convention.md and write file artifacts.
• In openspec mode, treat OpenSpec files on disk as canonical workflow state for continuation and recovery; never rely on conversation history.
• In none mode, return detected context only; write no SDD artifacts except the skill registry cache if required.
• Always persist testing capabilities in openspec/config.yaml testing: when mode is openspec.
• Always build .ospec/cache/skill-registry.cache.json.
• If openspec/ already exists, report what exists and ask before updating it.

Decision Gates

Input	Action
mode=openspec	Create/update openspec bootstrap files only.
mode=none	Return detected context only; write no SDD artifacts except the skill registry cache if required.
strict TDD marker/config found	Use that value.
no marker/config but test runner exists	Default strict_tdd: true.
no test runner	Set strict_tdd: false and explain unavailable.
existing code detected AND openspec/specs/ empty AND no baseline block	Activate brownfield branch: write baseline block, return next_recommended: sdd-baseline.
baseline block already present (any status)	Preserve it unchanged; if status is pending or partial, return next_recommended: sdd-baseline.
baseline.status: done	Brownfield branch does not activate; fall back to standard next_recommended logic.

Pre-Execution: Federated Bridge (target_dir + Multirepo Detection)

Run this resolution BEFORE any of the Execution Steps below. It resolves the base path and gates federated workspaces. No artifact is written until this resolution passes.

Step 0a — Resolve the base path from target_dir

• Read target_dir from the ## Parameters prompt block (the target_dir: <path> line), using the same injection pattern as ## Project Standards. There is no env var and no dynamic frontmatter field.
• When the ## Parameters block is absent or the target_dir key is missing, fall back to the current working directory (cwd). This is the backward-compatible default.
• When target_dir is present, fs.stat the path:
  • If it does not exist (ENOENT), STOP immediately and return status: blocked with a question_gate(invalid-path) describing the non-existent path. Do NOT create files at any location — the invalid-path gate fires before any artifact write.
  • If it exists, use it as the resolved base path; all artifact reads/writes are relative to it.

Step 0b — Multirepo container detection gate

After resolving a valid base path, scan its immediate children (depth-1 only, no recursion):

• If the resolved base path has no own .git (no .git of its own) AND has two or more (≥2) immediate children that each contain .git (directory OR file), treat it as a workspace container and STOP: return status: blocked with a question_gate listing exactly two options — federated (initialize as a federated workspace) and normal (initialize as a single repo). Never auto-select the federated path. This gate fires before any artifact write; no files are created until the user responds.
• If the base path has its own .git (single-repo), the container gate does NOT trigger and init falls through to the normal flow unchanged.
• If there are fewer than two children with .git (threshold is ≥2), the gate does NOT fire and init continues as a normal single-repo init.

Execution Steps

1. Inspect project files (package.json, go.mod, pyproject.toml, CI, lint/test config) and summarize stack/conventions.
2. Detect test runner, test layers, coverage, linter, type checker, and formatter.
3. Resolve Strict TDD from agent marker, openspec/config.yaml, detected runner fallback, or no-runner fallback.
4. Initialize persistence for the resolved mode.
5. Build .ospec/cache/skill-registry.cache.json using the skill-registry scan rules.
6. Persist testing capabilities and project context. 6b. Scale preset (openspec mode only): read the scale: <solo|team|enterprise> line from the ## Parameters prompt block (the orchestrator asks the user once at first init; absent → default team, ask nothing yourself). Write scale: {value} into openspec/config.yaml and materialize its preset:
   • solo: keep everything advisory — no extra blocks; routing prefers lite for trivial/small; clarify fires only on residual_ambiguity; 4R stays out of default route gates.
   • team (default): current defaults unchanged; the Change Collision Gate applies (it is always-on when other active changes exist); traceability trailers stay advisory.
   • enterprise: uncomment/write strict_tdd: true (when a runner exists), traceability: { trailers: required }, and mentorship: { mode: balanced }; keep 4R in the standard route gates; recommend declaring quality_gates: with on_fail: halt. On re-init with an existing scale: key, preserve it unchanged (same rule as the baseline block).
7. Brownfield branch (openspec mode only): if existing application code is detected outside openspec/, docs/, and dotfiles AND openspec/specs/ is empty AND openspec/config.yaml has no baseline block, write the baseline block with status: pending, empty domains_pending, domains_done, stale_domains, and last_checked: "". On re-init, if a baseline block already exists, preserve it unchanged.
8. Return the structured initialization envelope.

Output Contract

Return a structured result with these fields:

• status: success | blocked | partial
• executive_summary: one-sentence description of what was initialized
• artifacts: OpenSpec paths and registry paths written
• next_recommended: sdd-foundation for empty projects, otherwise sdd-explore or sdd-new
• risks: warnings about detected stack, Strict TDD status, or persistence setup
• skill_resolution: injected, fallback-registry, fallback-path, or none

Include project, stack, persistence mode, Strict TDD status, testing capability table, and saved paths in the detailed body.

References

• references/init-details.md — detection checklist, config skeleton, and output templates.
• ../_shared/openspec-convention.md — openspec layout and rules.