From spec-driven-delivery
Spec-driven delivery pipeline (brainstorming -> spec -> plan -> exec -> review). Triggers on substantial features, refactors, or anything that needs alignment before code. Merges Triora sprint workflows, Pillarworks SPRINT pattern, crash-repair-template architecture/decisions, and reality-check Spec ST-001 (acceptance-criteria-first). Distinct from bmad-planning (which is for greenfield).
How this skill is triggered — by the user, by Claude, or both
Slash command
/spec-driven-delivery:spec-driven-deliveryThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Five phases: BRAINSTORM (explore the problem), SPEC (write down the contract), PLAN (decompose into steps), EXEC (build per plan), REVIEW (verify against spec). The spec is the source of truth — every later phase references it.
Five phases: BRAINSTORM (explore the problem), SPEC (write down the contract), PLAN (decompose into steps), EXEC (build per plan), REVIEW (verify against spec). The spec is the source of truth — every later phase references it.
Use this for substantive work that isn't greenfield (greenfield → use bmad-planning instead).
Goal: explore the solution space before locking in.
Activities:
Output: a brainstorm doc (brainstorm.md or in chat). Short — under 1 page. Decision recorded.
Goal: write down the contract before building.
Spec template:
# Spec: <feature name>
**ID:** <project>-<NNN> (e.g., TRI-181, PW-S38, BRAIN-S28)
**Owner:** <person>
**Status:** draft | active | shipped | sunset
**Version:** 0.1.0
## Problem
(2-3 sentences in the user's words)
## Acceptance criteria
1. (testable assertion)
2. (testable assertion)
3. (testable assertion)
## Out of scope
(explicit non-goals)
## Constraints
- Compliance: PHI / GDPR / SOC2 / none
- Performance: <if any>
- Compatibility: <if any>
## Implementation notes
(short — link to the plan in TodoWrite or sprint doc for detail)
## Verification
(how we verify each acceptance criterion — manual / automated test / production check)
This is the contract. Subsequent phases reference it.
Goal: decompose the spec into ordered, sized steps.
Use TodoWrite. Each step:
Plan covers the whole spec before any step starts. Use methodology/plan-and-execute for the discipline.
Goal: build per plan.
methodology/plan-and-execute cadence (plan → execute → checkpoint → reflect)methodology/tdd-discipline (red-green-refactor)methodology/systematic-debuggingGoal: verify against spec.
methodology/code-review for the difffoundation/cross-model-verification for regex / async / migrationsOutput: spec marked shipped, optionally with a reflections section ("what we learned, what we'd do differently").
| Situation | Use |
|---|---|
| Greenfield feature, no existing code | bmad-planning (4-phase Brief/Model/Architect/Develop) |
| Substantial feature in existing code | spec-driven-delivery (this skill) |
| Refactor of existing module | spec-driven-delivery |
| Production bug fix | methodology/systematic-debugging |
| Tactical single-step task | None — just do it |
For specs that touch multiple Orryx Group entities (Triora + Pillarworks + Brain): the spec lives in orryx-knowledge/specs/cross-product/<spec-id>.md. Each affected entity's repo has a link back to the canonical spec, not a copy.
For single-entity specs: lives in the entity's repo under docs/specs/ or prompts/.
Spec is the single source of truth. Phase 3 (plan) references phase 2 (spec). Phase 4 (exec) references phase 3 (plan). Phase 5 (review) references phase 2 (spec). A spec change requires re-running phases 3-5 for affected scope; that's the cost of mid-flight scope change made visible.
Without a written spec: "what we're building" drifts during exec, review can't verify ("verify against what?"), and "done" becomes subjective.
| Source | Contribution |
|---|---|
| Triora sprint workflows (Sprint 181 active) | Sprint-shaped 5-phase cadence |
| Pillarworks SPRINT_*.md pattern | Per-sprint spec docs |
crash-repair-template architecture/{decisions,schemas}/ | ADR + schema-as-spec pattern |
| reality-check Spec ST-001 v0.1 | Acceptance-criteria-first framing |
methodology/bmad-planning — for greenfield work (use that instead)methodology/plan-and-execute — used inside Phase 3+4 for execution disciplinemethodology/tdd-discipline — used inside Phase 4 for new functionalitymethodology/code-review — used inside Phase 5delivery/sprint-plan (Wave 4) — for the sprint-shaped wrapper around specsGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.
npx claudepluginhub alexmclaren/orryx-knowledge --plugin spec-driven-delivery