Spectra Propose

Create a complete Spectra change proposal — from requirement to validated artifacts — in a single workflow.

Input: The argument after $spectra-propose is the requirement description. Examples:

• $spectra-propose add dark mode
• $spectra-propose fix the login page crash
• $spectra-propose improve search performance

If no argument is provided, the workflow will extract requirements from conversation context or ask.

Prerequisites: This skill requires the spectra CLI. If any spectra command fails with "command not found" or similar, report the error and STOP.

Steps

1. Determine the requirement source

   a. Argument provided (e.g., "add dark mode") → use it as the requirement description, skip to deriving the change name below.

   b. Plan file available:

   • Check if the conversation context mentions a plan file path (plan mode system messages include the path like <name>.md)
   • If found, check if the file exists at ``
   • If a plan file is found, use the AskUserQuestion tool to ask:
     • Option 1: Use the plan file
     • Option 2: Use conversation context
   • If conversation context has no relevant discussion, mention this when presenting the choice
   • If the user picks the plan file → read it and extract:
     • plan_title (H1 heading) → use as requirement description
     • plan_context (Context section) → use as proposal Why/Motivation content
     • plan_stages (numbered implementation stages) → use for artifact creation
     • plan_files (all file paths mentioned) → use for Impact section
   • If the user picks conversation context → fall through to (c)

   c. Conversation context → attempt to extract requirements from conversation history

   • If context is insufficient, use the AskUserQuestion tool to ask what they want to build

   From the resolved description, derive a kebab-case change name (e.g., "add dark mode" → add-dark-mode). Do not keep archive-style date prefixes in active change names. If the source name starts with YYYY-MM-DD-, strip that date prefix before running spectra new change; archived change names and directories are historical references, not active names to reuse.

   IMPORTANT: Do NOT proceed without understanding what the user wants to build.

2. Classify the change type

   Based on the requirement, classify the change into one of three types:

   Type	When to use
   Feature	New functionality, new capabilities
   Bug Fix	Fixing existing behavior, resolving errors
   Refactor	Architecture improvements, performance optimization, UI adjustments

   This determines the proposal template format in step 5.

3. Scan existing specs for relevance

   Before creating the change, check if any existing specs overlap:

   1. Use the Glob tool to list all files matching openspec/specs/*/spec.md
   2. Extract directory names as the spec identifier list
   3. Compare against the user's description to identify related specs (max 5 candidates)
   4. For each candidate (max 3), read the first 10 lines to retrieve the Purpose section
   5. If related specs are found, display them as an informational summary

   IMPORTANT:

   • If related specs are found, display them but do NOT stop or ask for confirmation — continue to the next step
   • If no related specs are found, silently proceed without mentioning the scan

4. Create the change directory

   spectra new change "<name>" --agent codex
   

   If a change with that name already exists, suggest continuing the existing change instead of creating a new one.

5. Write the proposal

   IMPORTANT — file path rules for the ## Impact section:

   • All file paths SHALL be written relative to the project root (e.g., src/lib/foo.ts, src-tauri/crates/core/src/bar.rs, docs/specs/specs/auth/spec.md).
   • Do NOT use relative fragments (e.g., parser/mod.rs, core/mod.rs) — preflight rejects them as non-anchored paths.
   • Do NOT wrap shell commands in backticks inside artifact text (e.g., `git mv a.rs b.rs`) — preflight's backtick extractor will otherwise mis-parse the command as a file reference.
   • When referring to a file without naming its concrete path, use descriptive prose (e.g., "Parser 入口檔") rather than a backticked path fragment.

   Get instructions:

   spectra instructions proposal --change "<name>" --json
   

   Generate the proposal content based on change type (see formats below), then write it via CLI:

   spectra new artifact proposal --change "<name>" --stdin <<'ARTIFACT_EOF'
   <proposal content>
   ARTIFACT_EOF
   

   If the command fails with a validation error, fix the content and retry.

   Use the following format based on change type:

   Feature

   ## Why
   
   <!-- Why this functionality is needed -->
   
   ## What Changes
   
   <!-- What will be different -->
   
   ## Non-Goals (optional)
   
   <!-- Scope exclusions and rejected approaches. Required when design.md is skipped. -->
   
   ## Capabilities
   
   ### New Capabilities
   
   - `<capability-name>`: <brief description>
   
   ### Modified Capabilities
   
   (none)
   
   ## Impact
   
   - Affected specs: <new or modified capabilities>
   - Affected code:
     - New: <paths to be created, relative to project root>
     - Modified: <paths that already exist>
     - Removed: <paths to be deleted>
   

   Bug Fix

   ## Problem
   
   <!-- Current broken behavior -->
   
   ## Root Cause
   
   <!-- Why it happens -->
   
   ## Proposed Solution
   
   <!-- How to fix -->
   
   ## Non-Goals (optional)
   
   <!-- Scope exclusions and rejected approaches. Required when design.md is skipped. -->
   
   ## Success Criteria
   
   <!-- Expected behavior after fix, verifiable conditions -->
   
   ## Impact
   
   - Affected code:
     - Modified: <paths that already exist>
     - New: <paths to be created, relative to project root>
     - Removed: <paths to be deleted>
   

   Refactor / Enhancement

   ## Summary
   
   <!-- One sentence description -->
   
   ## Motivation
   
   <!-- Why this is needed -->
   
   ## Proposed Solution
   
   <!-- How to do it -->
   
   ## Non-Goals (optional)
   
   <!-- Scope exclusions and rejected approaches. Required when design.md is skipped. -->
   
   ## Alternatives Considered (optional)
   
   <!-- Other approaches considered and why not -->
   
   ## Impact
   
   - Affected specs: <affected capabilities>
   - Affected code:
     - Modified: <paths that already exist>
     - New: <paths to be created, relative to project root>
     - Removed: <paths to be deleted>
   

6. Get the artifact build order

   spectra status --change "<name>" --json
   

   Parse the JSON to get:

   • applyRequires: array of artifact IDs needed before implementation
   • artifacts: list of all artifacts with their status and dependencies

7. Create remaining artifacts in sequence

   Loop through artifacts in dependency order (skip proposal since it's already done):

   a. For each artifact that is ready (dependencies satisfied):

   • Check if the artifact is optional: If the artifact is NOT in the dependency chain of any applyRequires artifact (i.e., removing it would not block reaching apply), it is optional. Get its instructions and read the instruction field. If the instruction contains conditional criteria (e.g., "create only if any apply"), evaluate whether any criteria apply to this change based on the proposal content. If none apply, skip the artifact and show: "⊘ Skipped (not needed for this change)". Then continue to the next artifact.

   • Get instructions:

     spectra instructions <artifact-id> --change "<name>" --json
     

   • The instructions JSON includes:

     • context: Project background (constraints for you - do NOT include in output)
     • rules: Artifact-specific rules (constraints for you - do NOT include in output)
     • template: The structure to use for your output file
     • instruction: Schema-specific guidance
     • outputPath: Where to write the artifact
     • dependencies: Completed artifacts to read for context
     • locale: The language to write the artifact in (e.g., "Japanese (日本語)"). If present, you MUST write the artifact content in this language. Exception: spec files (specs/*/.md) MUST always be written in English regardless of locale, because they use normative language (SHALL/MUST).

   • Read any completed dependency files for context

   • Generate the artifact content using template as the structure

   • Apply context and rules as constraints - but do NOT copy them into the file

   • Write the artifact via CLI (the CLI handles directory creation and format validation):

     For design or tasks:

     spectra new artifact <artifact-id> --change "<name>" --stdin <<'ARTIFACT_EOF'
     <content>
     ARTIFACT_EOF
     

     For specs (one command per capability):

     spectra new artifact spec <capability-name> --change "<name>" --stdin <<'ARTIFACT_EOF'
     <delta spec content>
     ARTIFACT_EOF
     

     If the command fails with a validation error, fix the content and retry.

   • Show brief progress: "✓ Created "

   b. Continue until all applyRequires artifacts are complete

   • After creating each artifact, re-run spectra status --change "<name>" --json
   • Check if every artifact ID in applyRequires has status: "done"
   • Stop when all applyRequires artifacts are done

   c. If an artifact requires user input (unclear context):

   • Use AskUserQuestion tool to clarify
   • Then continue with creation

8. Inline Self-Review (before CLI analysis)

   After creating all artifacts, scan them manually. Fix issues inline, then proceed to the CLI analyzer.

   Check 1: No Placeholders

   These patterns are artifact failures — fix each one before proceeding:

   • "TBD", "TODO", "FIXME", "implement later", "details to follow"
   • Vague instructions: "Add appropriate error handling", "Handle edge cases", "Write tests for the above"
   • Delegation by reference: "Similar to Task N" without repeating specifics
   • Steps describing WHAT without HOW: "Implement the authentication flow" (what flow? what steps?)
   • Empty template sections left unfilled
   • Weasel quantities: "some", "various", "several" when a specific number or list is needed

   Check 2: Internal Consistency

   • Does every capability in the proposal have a corresponding spec?
   • Does the design reference only capabilities from the proposal?
   • Do tasks cover all design decisions, and nothing outside proposal scope?
   • Are file paths consistent across proposal Impact, design, and tasks?

   Check 3: Scope Check

   • More than 15 pending tasks → consider decomposing into multiple changes
   • Any single task would take more than 1 hour → split it
   • Touches more than 3 unrelated subsystems → consider splitting

   Check 4: Ambiguity Check

   • Are success/failure conditions testable and specific?
   • Are boundary conditions defined (empty input, max limits, error cases)?
   • Could "the system" refer to multiple components? Be explicit.

   Check 5: Durable Handoff Review (run BEFORE the CLI analyzer)

   This change has to survive being parked or handed to another agent. Reject and fix any of the following:

   • File-path-only tasks: a task whose entire description is "edit file X" with no behavior, contract, or verification target. File paths are locator context — the task SHALL still describe what is observably true when complete.
   • Line-number-coupled instructions: design or tasks content that points to "line 42" / "the function on lines 80-95" as the only way to identify the work. Source line numbers drift; name the function, command, struct, or behavior instead.
   • Vague acceptance criteria: success conditions like "works correctly", "behaves as expected", "handles edge cases" without naming the observable behavior or the verification target (test name, CLI invocation, analyzer rule, manual assertion).
   • Missing scope boundaries on non-trivial work: design lacking explicit "in scope" / "out of scope" lines for any change that touches more than one subsystem or introduces new behavior. Trivial artifact-only edits MAY skip this; runtime, build, or tooling effects MUST NOT.

   Fix every failure inline using the existing context before running the CLI analyzer. If a failure cannot be fixed without new input from the user, surface it explicitly rather than papering over it.

Rationalization Table

9. Analyze-Fix Loop (max 2 iterations)

   1. Run spectra analyze <change-name> --json
   2. Filter findings to Critical and Warning only (ignore Suggestion)
   3. If no Critical/Warning findings → show "Artifacts look consistent ✓" and proceed
   4. If Critical/Warning findings exist: a. Show: "Found N issue(s), fixing... (attempt M/2)" b. Fix each finding in the affected artifact c. Re-run spectra analyze <change-name> --json d. Repeat up to 2 total iterations
   5. After 2 attempts, if findings remain:
      • Show remaining findings as a summary
      • Proceed normally (do NOT block)

10. Validation

    spectra validate "<name>"
    

    If validation fails, fix errors and re-validate.

11. Park the change and end the workflow

    Show summary:

    • Change name and location
    • List of artifacts created
    • Validation result

    Then unconditionally execute:

    spectra park "<name>"
    

    Inform the user that the change is parked and that running $spectra-apply <change-name> when ready will auto-unpark the change and start implementation.

    If you are currently in Codex Plan Mode, also remind the user to switch the session to normal mode before running $spectra-apply <change-name>. This is only a reminder: do NOT try to use ExitPlanMode or EnterPlanMode, do NOT ask whether to switch modes, and do NOT invoke apply.

    The propose workflow ENDS here. Do NOT invoke $spectra-apply. Do NOT call AskUserQuestion to ask whether to park or apply. This behavior is identical across Auto Mode, interactive mode, and any other agent mode — parking is unconditional and does not depend on AskUserQuestion availability or UI auto-accept settings.

Artifact Creation Guidelines

• Follow the instruction field from spectra instructions for each artifact type
• Read dependency artifacts for context before creating new ones
• Use template as the structure for your output file - fill in its sections
• IMPORTANT: context and rules are constraints for YOU, not content for the file
  • Do NOT copy <context>, <rules>, <project_context> blocks into the artifact
  • These guide what you write, but should never appear in the output
• Parallel task markers ([P]): When creating the tasks artifact, first read .spectra.yaml. If parallel_tasks: true is set, add [P] markers to tasks that can be executed in parallel. Format: - [ ] [P] Task description. A task qualifies for [P] if it targets different files from other pending tasks AND has no dependency on incomplete tasks in the same group. When parallel_tasks is not enabled, do NOT add [P] markers.

Guardrails

• Create all artifacts needed for implementation. Optional artifacts (those not in the applyRequires dependency chain) may be skipped if their inclusion criteria don't apply.
• Always read dependency artifacts before creating a new one
• If context is critically unclear, ask the user - but prefer making reasonable decisions to keep momentum
• If a change with that name already exists, suggest continuing that change instead
• Verify each artifact file exists after writing before proceeding to next
• NEVER write application code or implement features during this workflow
• NEVER skip the artifact workflow to write code directly
• NEVER reinterpret requirements by ignoring the proposal file
• NEVER invoke $spectra-apply — this workflow ends after artifact creation. The user decides when to start implementation
• If AskUserQuestion tool is not available, ask the same questions as plain text and wait for the user's response