spectre-evaluate

star 142

๐Ÿ‘ป | Architecture review + knowledge capture

Codename-Inc By Codename-Inc schedule Updated 5/24/2026

name: "spectre-evaluate" description: "๐Ÿ‘ป | Architecture review + knowledge capture" user-invocable: true

evaluate

Input Handling

Treat the current command arguments as this workflow's input. When invoked from a slash command, use the forwarded $ARGUMENTS value.

evaluate: Review architecture and capture learnings

Description

  • What โ€” Meta command that runs an architecture review in the background while you capture learnings from the session
  • Outcome โ€” Architecture review report + captured knowledge for future sessions

ARGUMENTS

$ARGUMENTS

Instructions

This command runs two activities in parallel:

  1. Architecture Review โ€” dispatched as a background Opus 4.6 subagent via Skill(spectre-architecture_review) (Claude slash route: spectre-architecture_review)
  2. Learn โ€” run directly by loading the spectre-learn skill

Step (1/2) - Dispatch Architecture Review

  • Action โ€” DispatchReview: Launch architecture review as a background subagent

    • Use the Task tool with subagent_type="general-purpose", model="opus", run_in_background=true
    • The subagent prompt should include the full architecture_review command instructions
    • Pass ARGUMENTS to the subagent as the review scope (feature description, paths, context)
    • If ARGUMENTS is empty: instruct the subagent to infer the review scope from recent work in the repository (e.g., git diff, recent commits)
    • Note the background task ID so the user can check results later
  • Output: Inform the user that the architecture review is running in the background and they'll be notified when it completes.

Step (2/2) - Capture Learnings

  • Action โ€” Learn: Load the spectre-learn skill and follow its workflow
    • Load skill: Skill(spectre-learn)
    • If ARGUMENTS is provided: use it as the learning topic
    • If ARGUMENTS is empty: the skill will analyze recent conversation to identify what's worth capturing
    • Follow the full learning workflow from the skill (categorize, propose, write, register)

Completion

When the learn flow completes:

  1. Check if the background architecture review has finished (read the output file)
  2. If finished: Present the architecture review report to the user
  3. If still running: Let the user know they can check back later or continue working
  • Action โ€” RenderFooter: Use Skill(spectre-guide) for Next Steps
Install via CLI
npx skills add https://github.com/Codename-Inc/spectre --skill spectre-evaluate
Repository Details
star Stars 142
call_split Forks 15
navigation Branch main
article Path SKILL.md
More from Creator
Codename-Inc
Codename-Inc Explore all skills →