From sdd-bootstrap
Interview-driven SDD bootstrap for project, feature, bugfix, or refactor work. Creates approved implementation-ready specifications and tasks from GitHub/GitLab issues or supplied requirements.
How this skill is triggered — by the user, by Claude, or both
Slash command
/sdd-bootstrap:sdd-bootstrap-interviewerThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Prepare work for implementation. This skill creates specifications and approved
references/api-contract-rules.mdreferences/architecture-review-checklist.mdreferences/architecture-rules.mdreferences/interview-question-bank.mdreferences/phase-quality-gates.mdreferences/task-splitting-rules.mdtemplates/AGENTS.template.mdtemplates/CLAUDE.template.mdtemplates/acceptance-tests.template.mdtemplates/adr.template.mdtemplates/ai-task.template.mdtemplates/c4-component.template.mdtemplates/c4-container.template.mdtemplates/c4-context.template.mdtemplates/ci-github.template.ymltemplates/ci-gitlab.template.ymltemplates/design.template.mdtemplates/gitlab-issue.template.mdtemplates/json-schema.template.jsontemplates/merge-request.template.mdPrepare work for implementation. This skill creates specifications and approved task contracts; it does not implement application code.
Codex:
Use the sdd-bootstrap-interviewer skill.
Mode: project | feature | bugfix | refactor
Source: <GitHub/GitLab issue URL or requirement text>
Claude Code:
/sdd-bootstrap:sdd-bootstrap-interviewer <project|feature|bugfix|refactor> <source>
feature, bugfix, and refactor modes, inspect related code, tests,
contracts, and established patterns. Parallel agents may be used only for
investigation and independent pre-implementation review.specs/<feature>/investigation.md exists, read it and carry all INV-xxx
and BL-xxx IDs forward into requirements and traceability.feature, bugfix, or refactor
modes, run investigate-codebase first and pass its outputs as context here.Open Questions; do not invent them.In feature, bugfix, and refactor modes, run
scripts/check-sdd-structure.sh (or .ps1) against the project root before
producing any specification artifacts. If the script reports any missing:
lines, run sdd-adopt (or perform its full process) to resolve every missing
item before continuing. Do not create specifications in a repository that lacks
the required SDD structure. Project-level constitution files (AGENTS.md,
CLAUDE.md) and CI/issue/PR templates are created by sdd-adopt; defer to it.
project: create the project constitution and first feature specification.feature: specify a new capability in an existing repository.bugfix: specify the observed behavior, expected behavior, regression test,
affected area, and smallest safe correction.refactor: specify a structural improvement that does not change observable
behavior. Requires specs/<feature>/investigation.md and
specs/<feature>/baseline-behavior.md; run investigate-codebase first if
they are absent. Acceptance criteria are expressed as BL-xxx behavior
equivalence.specs/<feature>/requirements.mdspecs/<feature>/design.mdspecs/<feature>/acceptance-tests.mdspecs/<feature>/tasks.mdspecs/<feature>/traceability.mddocs/adr/NNNN-<slug>.md for each new ADR (4-digit repository-wide sequence;
specs/<feature>/adr/ must not be created)CI/issue/PR templates are created by sdd-adopt based on detected host; do not
recreate them here.
For every generated task, propose a Risk: tier
(low | medium | high | critical) following
plugins/sdd-quality-loop/references/risk-classification-policy.md, with a
one-line Risk Rationale:. The agent proposes; the human confirms the tier at
approval. Derive Required Workflow: from the tier per
plugins/sdd-quality-loop/references/risk-gate-matrix.md:
low → test-after, medium → acceptance-first, high/critical → tdd.
For high/critical tasks, add the risk-derived Done When items the matrix
mandates (Red→Green evidence captured; independent review verdict recorded;
provenance with spec_revision; and for critical, an HMAC-signed evidence
bundle plus a second, distinct named approver). Leaving Risk: absent selects
legacy mode (no tier enforcement) and is reserved for pre-existing contracts —
do not use it to dodge a tier. check-risk rejects a high/critical task
that does not declare Required Workflow: tdd.
Generate every task with Approval: Draft and Status: Planned. Present the
specification and pre-implementation review to the human. Only a human may
change approval to Approved.
Do not approve tasks while requirements, design, contracts, acceptance criteria, scope, or important risks remain ambiguous.
If a valid SDD_SUDO flag file exists at the project root (see
plugins/sdd-quality-loop/references/sudo-mode-policy.md), the routine task
approval checkpoint auto-passes. Record
Approval: Approved (sudo <ISO8601 UTC>) in tasks.md and continue.
Sudo does not license approving ambiguous specifications: if requirements, design, contracts, acceptance criteria, scope, or important architecture/ security risks remain unresolved, keep them as Open Questions and do not auto-approve those tasks. Such decisions remain human-owned even under sudo. All deterministic gates apply; every check runs as normal.
Report generated files, open questions, risks, and the next draft task. Remind
the user that implementation starts with implement-task only after approval.
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
npx claudepluginhub aharada54914/sdd-forge --plugin sdd-bootstrap