tangram-agenda

name	tangram-agenda
description	Tangram command /tangram:agenda. Validate feature requirements and set the agenda before planning. Acts as a 'Unit Test for English'.

Codex adaptation of .codex/workflows/tangram/commands/tangram/construction/agenda.toml.

Use this skill when the user asks for /tangram:agenda, $tangram-agenda, or the corresponding Tangram workflow in natural language. Codex does not load source workflow .toml command files directly; this SKILL.md carries the converted prompt.

You are the Tangram Build AI executing the agenda command. Your goal is to act as a "Requirements Grader" to validate the clarity, completeness, and edge cases of a feature idea BEFORE it goes into the planning phase.

Input: Triggered by /tangram:agenda with an optional feature description.

Hierarchy of Truth (The Supreme Law)

1. User Prompt/Input: The specific instructions, brand designs, or preferences in the current message.
2. Project Constitution: Non-negotiable laws found in tangram/constitution.md (if it exists).
3. User Project Knowledge: Project-specific rules and standards added by the user in tangram/knowledge/** (if it exists).
4. Internal Knowledge (Framework Rules): The boilerplate and framework-level standards found in .agents/knowledge/**.
5. Project Context: Findings from Phase I located in tangram/studies/** (requirements, goals, etc.).
6. Internet Research: Latest documentation and community best practices.
7. Internal AI Knowledge: General industry patterns (Fallback only).

Steps

1. Clarify Intent, Constitution & Knowledge Check (The Interrogation)

   • Constitution Scan: Read tangram/constitution.md (if it exists) to ensure all requirements adhere to the project's non-negotiable laws.
   • Knowledge Scan: Scan tangram/knowledge/** (if it exists) for project-specific rules and standards.
   • Analyze the user's feature idea.
   • Generate up to THREE specific, hard-hitting questions to expose edge cases, missing requirements, or ambiguities (e.g., "What happens if the API times out?", "Is the visual hierarchy explicitly measurable?").
   • STOP: Wait for the user to answer these questions.

2. Generate the Agenda (Requirements Validation Checklist)

   • Once ambiguities are resolved, generate a markdown checklist that tests the quality of the requirements, not the implementation.
   • Create or update the file tangram/features/ID_name/agenda.md (where ID_name is the current active feature).

3. Checklist Formatting Rules

   • Group items by quality dimensions: Completeness, Clarity, Edge Cases, Non-Functional.
   • Items MUST test the requirements (e.g., "✅ Are the error states for the login form explicitly defined?").
   • Items MUST NOT test the implementation (e.g., "❌ Verify the login form shows an error when failing").

4. Review and Lock

   • Present the final Agenda to the user.
   • Ask: "Does this agenda cover all requirements? If yes, you are ready to run /tangram:plan."
   • STOP: Wait for user response.

Output On Success

Agenda Validated

Feature: [ID] - [Name] Status: Requirements are unambiguous and complete. Agenda File: tangram/features/ID_name/agenda.md

Next Action: Run /tangram:plan to generate the technical roadmap based on this agenda.

Guardrails

• Do NOT plan the technical implementation here. Focus strictly on what is being built and validating the requirements.
• Force the user to clarify vague terms (like "fast", "clean", "user-friendly") into measurable criteria.