epic: validate GitHub Projects as an optional planning projection

[sungjunlee]

Outcome

Validate the useful subset of GitHub Projects as an optional planning projection without making Projects a core dependency or a second source of task truth.

Questions to answer

• Can gh project and the GraphQL API reliably read/write iteration, priority, and roadmap fields for the target account/repository?
• Which operations are available through gh directly and which still require GraphQL?
• Can issue lifecycle remain canonical while Project fields are treated as planning metadata only?
• Is the operational cost lower than milestones + labels for the actual workflow?

Acceptance criteria

• A reproducible capability matrix covers authentication/scopes, item discovery, field discovery, iteration values, updates, and failure modes.
• One sandbox/pilot Project demonstrates optional iteration and priority projection from real Issues.
• No sprint, task resolver, or basic Issue → PR workflow requires Project access.
• Project writes are explicit and fail before local mutation when unavailable.
• The conclusion is evidence-backed: adopt a minimal optional profile or retain milestones + labels.
• No general Project abstraction is built before the pilot demonstrates repeated value.

Dependencies

May run after the authority contract; it must remain independent of mirror removal.

[coderabbitai]

🔗 Related PRs

#303 - feat: prove github and local tracker cycles end to end (#278) [merged]
#351 - docs: define GitHub-native authority contract [merged]
#353 - feat: make GitHub workflow mirrorless by default [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[sungjunlee]

Completion evidence: PR #355 and private pilot Project #5. Project #5 contains real Issues #345-#350 (6/6) and final Priority, Iteration, and Target Date read-back is 18/18. The CLI supports discovery, item add, simple-field creation, and one-field updates; GraphQL was required to create the Iteration field. Two planning cycles showed no repeated value: cycle 2 required four writes plus one read (45.8s), versus one milestone+labels read (4.0s), while sprint Plan already carried order. Invalid-auth and invalid-iteration writes both exited non-zero; before/after SHA-256 for Git status, active sprint, six canonical Issues, and six Project items/18 values were identical. Core absence acceptance passed 3/3; contract prose passed 8/8; Linux and Windows CI passed; independent final review reported no findings. Decision: retain milestones + labels + sprint Plan and add no Project profile, adapter, config, sync loop, or core dependency.