[typist] 🔤 Typist — Go Type Consistency Analysis

[github-actions[bot]]

Analysis of repository: github/gh-aw · 2026-07-03

Hi team! 👋 I swept every non-test .go file under pkg/ (942 files, 953 type definitions) hunting for two things: types we've accidentally defined twice, and places where we lean on any where a real type would keep us safer.

The good news first: there is essentially no accidental type duplication in this codebase. Every "duplicate name" I found turned out to be intentional — a type X = pkg.X re-export alias or a //go:build wasm platform variant. That's genuinely disciplined package design. 🎉

The real opportunity is weak typing. The codebase has fully migrated off interface{} (great!) but now threads a raw map[string]any frontmatter/config blob through 365 of 942 files, tearing it apart with 1,111 .(map[string]any) + 814 .(string) + 316 .([]any) assertions (1,708 total). A typed model (FrontmatterConfig) already exists but sits half-adopted. Finishing that migration plus adding a few named enum types (following the existing EngineName precedent) would remove the bulk of our runtime type-assertion risk. Nothing urgent — just high leverage.

Full Analysis Report

Duplicated Type Definitions — clean bill of health

• Total types analyzed: 953 · Duplicate-name clusters: 13 · Actionable duplicates: 0

All 13 clusters are intentional and need no action:

• 9 alias re-exports (type X = pkg.X, single source of truth): ActionYAMLInput, ActionPin, ActionPinsData, ContainerPin, SHAResolver (all pkg/actionpins/actionpins.go → pkg/workflow/action_pins.go); InputDefinition (pkg/types/input_definition.go:15 → pkg/workflow/inputs.go:18); SanitizeOptions (pkg/stringutil/sanitize.go:60 → pkg/workflow/strings.go:94); LogMetrics, ToolCallInfo (pkg/workflow/metrics.go → pkg/cli/logs_models.go).
• 3 //go:build platform variants (same package, !js && !wasm vs js || wasm — Go requires both): ProgressBar (pkg/console/progress.go:31), SpinnerWrapper (pkg/console/spinner.go:92), RepositoryFeatures (pkg/workflow/repository_features_validation.go:67; the only byte-identical redefinition, mandated by build constraints).
• 1 benign name collision: LogParser is an interface in pkg/workflow/agentic_engine.go:189 and a generic func type in pkg/cli/log_aggregation.go:37 — unrelated, different packages.

---

Untyped / Weakly-Typed Usages

• interface{} literals: 0 (already migrated to any 🎉)
• any usages: 3,625 total — map[string]any 2,171 (365 files), []any 415, []map[string]any 59, bare any params ~213
• Type assertions on any: 1,708 across 333 files

Category 1: the map[string]any config blob — 🔴 highest impact

map[string]any is used as a YAML/frontmatter config blob dismantled by hand. Assertion histogram: .(map[string]any) 1,111 · .(string) 814 · .([]any) 316 · .(bool) 146. A typed model already exists — FrontmatterConfig at pkg/workflow/frontmatter_types.go:301 (114-line struct, verified via Serena) — but the raw map is still threaded through hundreds of call sites in parallel.

Named-type opportunities

1. type FrontmatterMap map[string]any with accessor methods — biggest single win. The frontmatter/rawFrontmatter param appears 270× across 124 files. Seed points: pkg/parser/frontmatter_content.go:20 (FrontmatterResult.Frontmatter, verified), pkg/workflow/workflow_data.go:150 (RawFrontmatter), pkg/cli/mcp_workflow_scanner.go:21. Add GetString/GetMap/GetSlice/GetBool methods so 124 files stop asserting by hand.

2. Finish the parse*Config(outputMap map[string]any) migration — 48 functions share this signature (e.g. pkg/workflow/publish_artifacts.go:53 parseUploadArtifactConfig, pkg/workflow/dispatch_repository.go:30, update_project.go:36, assign_to_user.go:19, hide_comment.go:19, ...), each re-implementing pull→assert→default.

3. tools map[string]any blob — 64×, e.g. pkg/workflow/claude_tools.go:20 expandNeutralToolsToClaudeTools(tools map[string]any) map[string]any.

Assertion hotspots (migrate first): pkg/workflow/frontmatter_extraction_security.go (70 — unmarshal into typed NetworkConfig/SandboxConfig), repo_memory.go (47), safe_outputs_config.go (39), engine.go (39), pkg/parser/import_field_extractor.go (34).

Leave alone (idiomatic any): true JSON/YAML decode boundaries, template payloads, and genuine polymorphic Default any fields (pkg/types/input_definition.go:18).

Category 2: missing semantic types on enum-like constants — 🟡 medium

The codebase shows the good pattern once (pkg/constants/engine_constants.go:13 type EngineName string). These bare-string/int groups are effectively enums that didn't get it:

Enum candidates

• MCP server type — pkg/types/mcp.go:13 Type string (verified), validated in pkg/parser/mcp.go:22 and compared to literals ~12× (mcp_config_custom.go:104,119). → type MCPServerType string.
• MCP auth type — pkg/types/mcp.go:33 (only github-oidc). → type MCPAuthType string.
• InputDefinition.Type — pkg/types/input_definition.go:19 (5-value enum, switched at build_input_schema.go:44-49). → type InputType string.
• Registry — pkg/cli/mcp_registry_types.go:108-115 StatusActive/Inactive, ArgumentTypePositional/Named (bare). → type ServerStatus string, type ArgumentType string.
• Objective labels — pkg/github/label_objective_mapping_constants.go:21+ (24 bare-string consts). → type ObjectiveLabel string.
• Safe-output kinds — raw literals at pkg/cli/outcome_eval.go:125,391 (noop/missing_tool/missing_data/report_incomplete), also logs_metrics.go:380,510,641. → type SafeOutputKind string.
• Ports — pkg/constants/constants.go:105-135 (DefaultMCPGatewayPort=8080, MinNetworkPort=1, MaxNetworkPort=65535, ..., bare int). → type Port int.

Leave alone: arithmetic constants in pkg/workflow/time_delta.go:489-501 and byte-size limits in pkg/workflow/compiler.go:25-38; env-var-name strings in engine_constants.go:213-364.

---

🎯 Recommendations (impact-per-effort)

1. High — FrontmatterMap + accessors. Define the type, retype FrontmatterResult.Frontmatter & WorkflowData.RawFrontmatter, migrate the 5 hottest decoder files. Incremental; removes the bulk of 1,700+ assertions.
2. Medium — named enum types. Start with MCPServerType, InputType, SafeOutputKind. Compile-time safety against typos.
3. Low — finish parse*Config standardization onto the accessors. Mechanical; less boilerplate.

Checklist: ☐ FrontmatterMap + methods + tests ☐ retype seed fields ☐ migrate top-5 decoder files ☐ MCPServerType/MCPAuthType/InputType ☐ SafeOutputKind/ArgumentType/ServerStatus ☐ type Port int ☐ run tests per package.

Metadata

942 files analyzed · 953 types · 13 duplicate clusters (0 actionable) · 3,625 any + 1,708 assertions · Method: pattern sweep + Serena semantic verification (activate_project, find_symbol) · 2026-07-03

References: §28659147298

Generated by ✍️ Typist - Go Type Analysis · 552.8 AIC · ⌖ 16.4 AIC · ⊞ 8.8K · ◷

• expires on Jul 4, 2026, 4:20 AM UTC-08:00