[Schema Consistency] Schema Consistency Check — 2026-06-15

[github-actions[bot]]

🔍 Schema Consistency Check — 2026-06-15

Summary

• Inconsistencies Found: 3 (1 HIGH, 2 LOW)
• Categories Analyzed: Schema ↔ Parser ↔ Documentation ↔ Workflows
• Strategy Used: strategy-25 (Engine sub-object key coverage) + strategy-26 (Engine sub-key doc coverage)
• Selection Mode: Proven strategy (day-of-year 166 mod 10 = 6, ≤ 6)
• New Strategy Added: Yes — strategy-26 (engine sub-key documentation coverage)

Critical Issues

🔴 HIGH — engine.extensions is a triple gap (parser ↔ schema ↔ docs)

The Pi engine supports an engine.extensions array, but it is only implemented in the parser. It is absent from both the JSON schema and the documentation, so any workflow that uses it is rejected by strict schema validation despite being a working runtime feature.

Layer	State
Parser	✅ Reads engineObj["extensions"] at pkg/workflow/engine.go:469 (handles both []any and []string)
Consumer	✅ Consumed at pkg/workflow/pi_engine.go:202 to emit pi install <extension> steps
Schema	❌ engine_config has 6 oneOf branches (union = 28 props), all additionalProperties: false; none declares extensions
Docs	❌ Not in docs/.../reference/engines.md nor auto-generated frontmatter-full.md (only the unrelated safe-output key allowed-extensions appears)

Impact: Because every engine_config branch is additionalProperties: false, a frontmatter such as

engine:
  id: copilot
  extensions: ["my-ext"]

will fail schema validation, even though the compiler would otherwise emit the install steps correctly. The feature is effectively unreachable through the validated path and invisible to users.

Evidence

• pkg/workflow/engine.go:468-488 — extraction of engine.extensions
• pkg/workflow/pi_engine.go:175,202 — pi install step generation
• pkg/parser/schemas/main_workflow_schema.json:11419 — engine_config $def; 6 oneOf branches, union of properties does not include extensions
• docs/.../reference/engines.md — no extensions mention (only "agent file" / "bare command" extension references)

Suggested fix: Add extensions (array of strings) to the relevant engine_config branch(es) — at minimum the object branch that carries id/provider for the Pi engine — and document it in engines.md.

Schema ↔ Parser Mismatches (low priority)

🟡 LOW — dispatch-repository (dash) alias accepted by parser but not in schema

The parser accepts both spellings for the safe-output key:

// pkg/workflow/dispatch_repository.go:37-41
if configData, exists = outputMap["dispatch_repository"]; !exists {
    if configData, exists = outputMap["dispatch-repository"]; !exists {
        return nil
    }
}

The schema only declares dispatch_repository (underscore) at main_workflow_schema.json:9120, under a safe-outputs object with additionalProperties: false. A workflow using the dash spelling works at compile time but fails schema validation.

Suggested fix: Either add dispatch-repository as a schema alias property, or drop the dash alias from the parser to keep a single canonical spelling.

Schema Naming Consistency (low priority)

🟡 LOW — dispatch_repository is the sole snake_case key among kebab-case safe-output siblings

All other safe-outputs keys use kebab-case (create-pull-request, create-issue, add-comment, ...); dispatch_repository is the lone snake_case key. Cosmetic, but inconsistent. Aligning to dispatch-repository (and treating the underscore form as the alias) would resolve this and the item above together.

Recommendations

1. (HIGH) Add extensions to the Pi engine_config schema branch and document it in engines.md, OR explicitly gate/remove the parser support if the feature is not intended to be public.
2. (LOW) Reconcile the dispatch_repository / dispatch-repository spelling between parser and schema — pick one canonical form and declare any alias in the schema.
3. (LOW) Standardize the safe-output key on kebab-case for consistency.

Strategy Performance

Metric	Value
Strategy used	strategy-25 (engine sub-object key coverage) + strategy-26 (new)
Findings	1 HIGH (re-confirmed), 2 LOW (re-confirmed)
Effectiveness	HIGH
Should reuse	Yes — engineObj-key-vs-branch-union diff remains the highest-signal probe

Next Steps

• Add engine.extensions to schema (engine_config branch) and engines.md
• Reconcile dispatch_repository vs dispatch-repository (parser ↔ schema)
• Standardize safe-output key casing
• Future run: apply the engineObj-style coverage diff to safe-outputs entry sub-objects and tools.* sub-objects (note: those parsers use struct unmarshalling rather than map-key reads, so a different probe is required)

References: §27529660516

Generated by ✅ Schema Consistency Checker · 112.6 AIC · ⌖ 31.4 AIC · ⊞ 8.9K · ◷

• expires on Jun 15, 2026, 11:10 PM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by Schema Consistency Checker.

A newer discussion is available at Discussion #39506.