Repository navigation
Description
Problem
Squad's distributable gh-aw workflow now supports /squad cast, /squad research, /squad plan, /squad plan accept, and plan revision. The Aspiregregator modernization exercise proved this can cast a credible specialist team, research a real repository, and produce a substantial backlog with acceptance criteria and native sub-issue links.
It also exposed an ontological gap: the current flow moves directly from research to issue creation without an explicit model for triage, decisions, initiatives/epics, milestones, implementation-ready work, sizing, dependency validation, or staged acceptance.
The result is a strong research-derived backlog, but not necessarily an executable SDLC plan.
Reference exercise: bradygaster/Aspiregregator#8 and its generated children #12–#25.
Goal
Evolve the Squad planning workflow into a reusable planning capability for customer repositories and downstream automation.
Target lifecycle:
Intent
→ Cast
→ Research
→ Triage
→ Decision gate
→ Program plan
→ Implementation plan
→ Plan validation
→ Accept scope
→ Accept implementation plan
→ Activate execution
This capability belongs to Squad and must remain independently useful and customizable.
Core ontology
| Concept | Purpose | Preferred GitHub representation |
|---|---|---|
| Intent | Authoritative desired outcome and constraints | Root issue |
| Finding | Evidence discovered during research | Structured research comment/artifact; not automatically work |
| Disposition | Triage decision for a finding | Structured plan data and resulting work classification |
| Decision | Human choice required before planning can safely continue | Issue or explicit decision gate |
| Initiative | Overall outcome-bearing body of work | Root issue / native issue type when available |
| Epic | Coherent capability or workstream | Native parent issue |
| User story / feature | User-observable increment | Issue beneath an epic |
| Task | PR-sized implementation unit | Leaf issue |
| Milestone | Independently demonstrable delivery outcome | GitHub milestone |
| Dependency | Execution ordering constraint | Native blocked-by relationship |
| Size | Relative implementation effort/uncertainty | GitHub Project field when available; structured Size: XS–XL fallback |
| Acceptance | Approval of scope, implementation shape, or activation | Separate explicit commands/transitions |
Design principles
- Preserve the initiating intent through every phase; after-the-fact triage must not reconstruct intent from generated issues alone.
- Research produces evidence, not an automatic backlog.
- Triage decides what becomes work, what needs a decision, and what is excluded.
- Program planning and implementation planning are separate products.
- Native GitHub hierarchy, dependencies, milestones, issue types, Projects fields, and state should be preferred over labels.
- Labels are a fallback for dimensions GitHub cannot otherwise represent and must include names, descriptions, and intentional color families.
- Sizing is relative planning metadata, not elapsed-time prediction. Prefer a Project single-select field; use
Size: XS/S/M/L/XLin structured issue metadata when no project is configured rather than creating repository labels by default. - Acceptance must validate postconditions before creating or activating work.
- Commands and output contracts must support persistent and ephemeral Squad execution models.
Proposed command surface
/squad research
/squad triage
/squad plan program
/squad plan implementation
/squad plan validate
/squad plan accept scope
/squad plan accept implementation
/squad plan activate
/squad plan revise <feedback>
Backward compatibility for /squad plan and /squad plan accept should be deliberate and documented.
Acceptance criteria
- The lifecycle ontology and structured artifacts are documented and machine-validated.
- Research findings can be triaged against the original intent before issues are created.
- Program plans can create initiatives/epics, outcome milestones, and user-story/feature groupings.
- Implementation plans can create PR-sized leaf work with real dependencies, acceptance criteria, and relative sizes.
- Plan validation detects unresolved temporary IDs, missing mappings, invalid hierarchy, dependency cycles, oversized work, missing decisions, and incomplete metadata.
- Scope acceptance, implementation acceptance, and execution activation are distinct operations.
- GitHub-native representations are used when available, with predictable fallbacks when issue types, Projects, or permissions are unavailable.
- Customers can customize ontology and policy without forking core workflow logic.
- Downstream automation can consume accepted planning contracts without Squad depending on a specific executor.
- Aspiregregator is captured as an end-to-end regression fixture or documented acceptance scenario.
Out of scope
- Implementing downstream execution, review, merge, or completion workflows.
- Making Squad state persistence mandatory for gh-aw runs.
- Automatically implementing accepted work.
- Requiring every customer repository to configure GitHub Projects.
Size
XL — epic; decomposed into independently deliverable children below.
Activity
fix(gh-aw): resolve review blockers for SDLC lifecycle PR
fix(gh-aw): resolve review blockers for SDLC lifecycle PR
feat: Evolve Squad planning into executable SDLC lifecycle (#1648) (#…