Repository navigation
Feat/4913 workflow mcp inventory - #5054
Conversation
Assisted-by: OpenAI Codex (model: GPT-6, autonomous)
Assisted-by: OpenAI Codex (model: GPT-6, autonomous)
|
@mnriem Thanks for the guidance on keeping workflow MCP contributions small and focused! Just to clarify the scope of #5054: this is a foundational inventory-only PR covering the 29 currently registered workflow CLI commands. The ~924 added lines are primarily explicit per-command metadata and parity/regression tests. It introduces no MCP adapters or runtime behavior changes. The goal is to establish a single inventory and test baseline so future contributions can be smaller and easier to review. I'll follow your suggested approach for the next PRs: one workflow command/adapter at a time, with its own typed contract and focused tests. Happy to adjust the foundational scope if you'd prefer a different split. |
There was a problem hiding this comment.
🟢 Approval recommended
The inventory introduces no runtime exposure, and the remaining feedback is non-blocking.
2 open issues
What changed in this PR
Adds hierarchy-owned MCP inventory for all 29 workflow CLI leaves without exposing new tools or changing workflow behavior.
Changes:
- Records unavailable dispositions, reasons, capabilities, and network metadata.
- Adds CLI parity checks, invalid-record tests, and discovery/effect coverage.
| File | Description |
|---|---|
tests/specify_cli/workflows/test_mcp.py |
Validates inventory parity, metadata, and unchanged workflow exposure. |
src/specify_cli/workflows/_mcp.py |
Defines inventory records for every workflow CLI leaf. |
Review effort: Max
Assisted-by: OpenAI Codex (model: GPT-6, autonomous) Signed-off-by: Nishanth <bnishanthtpm06@gmail.com>
|
Addressed both Copilot review comments and pushed the corrections.
Validation: 73 tests passed on Linux, with Ruff and diff checks passing. @mnriem, the PR is ready for another look whenever convenient. Thanks! |
|
Thank you! |

Description
Adds a foundational workflow MCP inventory covering all 29 registered
specify workflowCLI leaves, including nested command groups.Each command is explicitly classified as
unavailable, with a concrete disposition reason and source-audited capability and network-access metadata. No MCP adapters, server registration, or workflow behavior changes are introduced.Parity tests inspect the actual registered Typer hierarchy and reject missing, stale, duplicate, and invalid inventory records. Additional tests protect execution metadata and verify that existing MCP discovery and the legacy command catalog remain unchanged.
Available records fail closed until operation-specific contract evidence exists. Separate synthetic tests cover MCP tool discovery cardinality and schema shape without claiming contract correspondence.
Refs #4913.
Testing
git diff --checkpassed.specify workflow --helpoutput matched the clean baseline.The full repository test suite was not run.
Repository template checklist:
uv run specify --help— equivalent workflow help comparison completed; exact command not run.uv sync && uv run pytest— selected regression suites passed using existing environments; exact command not run.AI Disclosure
AI disclosure: OpenAI Codex (GPT-6.1 sol, autonomous within contributor-authorized scope; reasoning-effort medium) performed source inspection, effect auditing, implementation, adversarial checks, test execution, and drafting of this PR description. An independent AI-assisted review identified test gaps that were subsequently addressed. This disclosure does not claim human line-by-line review.