/gate-check — Gate Check

Category: Governance · Primary agent: project-planning-specialist

Runs a phase gate: checks exit criteria for the current stage and returns PASS / CONCERNS / FAIL / NOT ASSESSED.

Arguments: $ARGUMENTS

Purpose

Run a phase gate: check, criterion by criterion, whether the evidence for leaving the current stage exists in the project's files, collect specialist verdicts, and return an overall verdict with blockers, owners and next actions. A gate never passes by default; missing evidence means NOT ASSESSED.

Use it when a project wants to move from one stage to the next (concept, design, build, estimate, validate, operate), or before a milestone review.

Do not use it when the question is release approval for a model (/model-risk-review, /model-release).

Before you start

1. Check required inputs. For each input below record FOUND or ABSENT (never 'assumed'):
   • Stage artifacts
   • studio-config.yaml If a required input is ABSENT, that part is NOT ASSESSED — NO DATA: do not estimate or invent it; name what is missing and which skill produces it. If all required inputs are absent, stop and say so.
2. Check confidentiality. Identify the data classification level of the inputs (.claude/docs/confidentiality-and-ai-use.md). Do not proceed with restricted or confidential data in a way the policy does not allow.
3. Collaboration. Follow .claude/docs/collaboration-protocol.md: ask questions first, present options, get approval before writing files, and update production/session-state/active.md after each approved section.

Agents

• project-planning-specialist — primary agent; coordinates the work and integrates results
• economic-reasoning-specialist — Frames the economic question and tests whether results make economic sense
• modeling-methodology-specialist — Chooses methods and architecture across model families, sets code and data standards, and resolves technical trade-offs between traditional, structural and AI approaches
• model-risk-governance-specialist — Defines model risk levels and validation depth, enforces builder-validator independence, governs AI/LLM use and recommends release or block

Run these agents as subagents (the Task tool), in parallel where the steps are independent. Subagents only write to the paths named in your instructions.

Workflow

1. Read the gate definition in .claude/docs/phase-gates.md
2. Check every criterion against actual files; mark FOUND or ABSENT
   • Cite the file path for each FOUND item; for lego models include the manifest check status
3. Ask the economic-reasoning, methodology and model-risk specialists for a verdict on their criteria
4. Return an overall verdict with blockers, owners and next actions

Checkpoints with the user

• Before the verdict is recorded: review blockers and owners

Common pitfalls

• Passing on promises rather than files
• Criteria marked FOUND without a file reference
• Blockers without owners

Output

production/gates/<stage>-<date>.md

The gate record lists each criterion with FOUND or ABSENT and its evidence path, specialist verdicts, the overall verdict, blockers, owners and next actions.

Gate / verdict

• Verdict 'NOT ASSESSED' if required evidence is absent — never a pass by default
• Every FOUND criterion cites a file

Suggested next steps

• /milestone-review