Task 7: AgentFrontmatter Enum Models

Task 7: AgentFrontmatter Enum Models

Status: NOT STARTED Agent: @python-cli-architect Dependencies: Task 2 Priority: 2 Complexity: S Accuracy Risk: High

Context

Existing AgentFrontmatter model (lines 1167-1235) lacks enum validation for model, permissionMode, and memory fields. Need to add enum types for strict validation.

Objective

Enhance AgentFrontmatter Pydantic model with enum validation for agent-specific fields.

Required Inputs

• Architecture spec: ./architect-plugin-linter.md lines 267-312 (Agent enum schema)
• Official docs: https://docs.anthropic.com/en/docs/claude-code/sub-agents.md (cite)
• Current AgentFrontmatter model: plugin_validator.py lines 1167-1235

Requirements

1. Create AgentModel StrEnum with values: SONNET, OPUS, HAIKU, INHERIT
2. Create AgentPermissionMode StrEnum with 6 permission mode values
3. Create AgentMemory StrEnum with values: USER, PROJECT, LOCAL
4. Modify AgentFrontmatter model to use these enum types
5. Change model: str | None to model: AgentModel | None
6. Change permissionMode: str | None to permissionMode: AgentPermissionMode | None
7. Add memory: AgentMemory | None field if not present

Constraints

• MUST cite official schema URL in enum docstrings
• MUST preserve all existing AgentFrontmatter fields unchanged
• MUST use exact enum value casing from official docs
• MUST NOT break existing agent frontmatter parsing

Expected Outputs

• Modified file: plugins/plugin-creator/scripts/plugin_validator.py (lines 1167-1235 region)
• 3 new StrEnum classes added
• AgentFrontmatter model fields updated to use enums
• Docstrings cite official schema URL

Acceptance Criteria

1. AgentModel enum contains exactly 4 values (sonnet, opus, haiku, inherit)
2. AgentPermissionMode enum contains all 6 permission modes
3. AgentMemory enum contains 3 memory scope values
4. AgentFrontmatter.model field type is AgentModel | None
5. Existing agent files still validate correctly
6. Invalid enum values rejected with Pydantic validation error

Verification Steps

1. Parse existing agent .md files from codebase, verify no validation errors
2. Create test agent with invalid model value "gpt-4", verify rejection
3. Create test agent with invalid permissionMode, verify error suggests valid values
4. Run mypy --strict on modified file
5. Verify backward compatibility with existing tests

CoVe Checks

Accuracy Risk: High (must match official agent schema exactly)

• Key claims to verify:

  • All enum values match official documentation exactly
  • Enum value casing is correct (lowercase vs camelCase)
  • No valid enum values are missing

• Verification questions:

  1. Are model values lowercase or capitalized in official schema?
  2. Is permissionMode camelCase or snake_case in official schema?
  3. Are there additional permission modes beyond the 6 listed?
  4. Is memory field actually supported in current Claude Code version?

• Evidence to collect:

  • Fetch official agent schema: WebFetch("https://docs.anthropic.com/en/docs/claude-code/sub-agents.md")
  • Cross-reference all enum values with official list
  • Test with actual agent files from this repository
  • Check Claude Code version compatibility

• Revision rule:

  • If official schema has different enum values, update to match official
  • If memory field is not in official schema, mark as experimental in docstring
  • Document any version-specific enum values

Can Parallelize With: Task 4 (HookConfig), Task 5 (MCPConfig), Task 6 (LSPConfig) Reason: Independent schema enhancements Handoff: Provide model diff, verification outputs, official schema comparison