feat(workflow): formal test suite for intent-attribution-agent-governance spec (P1–P11 + safety invariant)

[Copilot]

Formalizes specs/intent-attribution-agent-governance.md as an executable Go test suite, encoding 11 core predicates and a cross-cutting safety invariant. Since the real attribution resolver doesn't exist yet (Phase 2+), the file ships self-contained stubs that directly encode the spec semantics and are marked for replacement.

New file: pkg/workflow/intent_attribution_governance_formal_test.go

Stub types (replace with real impl when Phase 2+ lands):

• pullRequestData — resolver inputs (explicit intent, closing issues, labels)
• formalIntentRecord — normalized output with Status, Source, Key, Risk, Priority, Domains
• formalExecutionPolicy — governance output with autonomy, write scope, approval/merge/attempt constraints

Stub functions encoding the spec directly:

• formalResolve — attribution-resolution order: explicit metadata → single closing issue → multi-issue ambiguous → label fallback → unlinked
• formalSafestPolicy — fail-closed floor: propose_only / none / human_approval_required: true / auto_merge_allowed: false / max_attempts: 1
• formalDerivePolicy — indeterminate statuses (unlinked, ambiguous, suggested, unmapped) always return formalSafestPolicy()
• formalResolveRisk — explicit risk wins; else derived from domain+priority (security+critical→high, production→high, infrastructure→medium, documentation→low, unknown→unknown)
• formalPolicyPrecedenceLevel — agent_request(0) < workflow(1) < intent(2) < repository(3) < organization(4)

Predicates tested:

Test	Predicate
P1	Explicit intent overrides ambiguous closing-issue candidates
P2	Single closing issue → mapped / closing_issue
P3	2+ closing issues → ambiguous, order-independent (3 orderings)
P4	No closing issues + labels → artifact_label fallback
P5	No source at all → unlinked
P6	ambiguous ≠ mapped for authorization purposes
P7	Indeterminate statuses always yield safest policy (4 cases)
P8	Identical inputs always produce identical policy (determinism)
P9	Risk derivation order with explicit override and 6 domain cases
P10	Precedence ordering including non-adjacent org > agent_request check
P11	Unlinked/ambiguous attribution never grants elevated autonomy, auto-merge, or max-attempts > 1
SAFETY	Ambiguous always fails closed unconditionally

• Fixes [formal-spec] intent-attribution-agent-governance.md — Formal model & test suite — 2026-07-30 #49170

[github-actions]

Warning

threat detection engine error
The threat detection engine encountered an error and could not complete analysis. This is a tooling failure, not a security finding.

Details

The threat detection engine failed to produce results.

Review the workflow run logs for details.

✅ Contribution Aligned

Thanks for formalizing the intent-attribution governance spec into an executable test suite! This is exactly what Phase 1 needs — a crystal-clear set of predicates (P1–P11) plus the safety invariant encoding the spec semantics.

What looks great:

• Comprehensive predicate coverage — all 11 core predicates plus the cross-cutting safety invariant are present and well-documented
• Clear stub marking — stubs are explicitly marked for replacement in Phase 2+, making it easy for implementers to see what's placeholders and what's spec-defined behavior
• Determinism and fail-closed guarantees — safety-critical checks like ambiguous attribution always yielding the safest policy are front and center
• Well-organized test table — predicates are easy to scan and understand at a glance

The PR is focused, documented, and ready for merge. Looking forward to Phase 2+ when the real resolver ships and these stubs get replaced with live implementations! 🎯

Generated by ✅ Contribution Check · auto · 55.7 AIC · ⊞ 8.7K · ◷

[github-actions]

PR Triage

Category: test - Risk: low - Score: 40/100

Impact	Urgency	Quality
16	8	16

Recommended action: batch_review (batch: spec-docs-20260730)

New self-contained Go test file (+476/-0) encoding governance spec predicates with explicit stub types marked for future replacement. No production code touched, draft, CI pending. Batch with #49177 and #49174 for combined review. Confirm stubs are clearly marked for Phase 2+ replacement before merge.

Generated by 🔧 PR Triage Agent · auto · 46.2 AIC · ⌖ 7.96 AIC · ⊞ 7.9K · ◷