enhance(spec-system-map): add deeper repo evidence pass before create

[sungjunlee]

Context

spec-system-map create currently asks the agent to read bounded signals, but the contract is loose enough that a generated map can feel like it only skimmed the repo. The desired behavior is deeper code investigation before writing, without turning spec/system-map.md into a module inventory.

Desired change

Add a Repo Evidence Pass to spec-system-map create mode.

The pass should inspect enough code reality to understand system shape:

• entrypoints and command surfaces
• package/config scripts
• runtime boundaries and external systems
• tests that reveal intended behavior
• storage/state surfaces
• recent commit/sprint evidence when available

The final spec/system-map.md should remain short. Evidence details should appear in the run report as Evidence Read / Evidence Missing, not as exhaustive map content.

Acceptance Criteria

• skills/spec-system-map/SKILL.md create mode includes a Repo Evidence Pass before drafting.
• The pass names the evidence classes agents should inspect.
• Completion output for create mode includes concise Evidence Read and Evidence Missing bullets.
• Quality checks continue to reject module inventories, endpoint lists, runbook commands, and temporary TODOs.
• Existing eval prompts are updated or extended to catch shallow brownfield maps.

[coderabbitai]

🔗 Related PRs

#106 - docs(backlog-charter): create-mode interview + signals + seed-Decisions guidance (#93 #94 #96) [merged]
#174 - feat(backlog-triage): add spec-aware Decision Review [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]

Implementation note after scope clarification: keep this Markdown-first. The first pass should update skills/spec-system-map/SKILL.md, templates/eval prompts, and reporting expectations. Do not add a new repo-survey JS script unless a later issue explicitly calls for it; the Repo Evidence Pass is an agent checklist plus concise evidence report, not a new authoritative analyzer.