Add skill forge lab, /mobius-profile skill, and ideas backlog#6
Merged
AaronGoldsmith merged 8 commits intomainfrom Mar 15, 2026
Merged
Add skill forge lab, /mobius-profile skill, and ideas backlog#6AaronGoldsmith merged 8 commits intomainfrom
AaronGoldsmith merged 8 commits intomainfrom
Conversation
…flow - Add "when Mobius is worth it" section with honest trade-off framing - Add prerequisites section with mandatory vs optional API keys - Show all three bootstrap options (API, Claude Code, scout) upfront - Add Trade-off column to "Why Mobius?" table - Replace architecture file listing with orchestrator flow diagram - Add context and caveats to cost table (token range, date, local embeddings) - Note env-overridable vs code-only config options - Remove duplicate `mobius train` command entry Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ount - Fix cost date from March 2025 to March 2026 - Add `mobius agent export` to commands list (exists in CLI but was undocumented) - Fix scout demo showing 4 agents when --count 5 was requested Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix scout bootstrap: "free API cost" → "~$0.50" (uses Opus API) - Add scout row to cost table - Clarify Claude Code note: CLI commands still require API keys - Standardize subscription naming to "Pro/Team" everywhere - Clarify Anthropic needed for default judge panel (not just "recommended") - Change "rate limiting" to "concurrency control" (semaphore in swarm.py) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Covers Pro, Max, and Team without needing updates as plans change. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…uirements, and trade-off wording Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
First two-phase competition: 6 agents designed Claude Code skills, then 6 testers validated them against the live DB. The generate→test pattern proved more valuable than any individual skill produced. Relates to #5 (mobius improve self-diagnosing loop) Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Adds documentation and a new Claude Code skill focused on analyzing a single Mobius agent’s performance, plus updates the project README to better explain when/how to use Mobius and how to bootstrap agents.
Changes:
- Added a lab write-up documenting the “Skill Forge” two-phase (generate → functional test) competition pattern.
- Added a new
/mobius-profileClaude Code skill (SKILL.md +show_profile.py) to display agent match history and recommend challengers. - Added
ideas.mdbacklog and refreshed README positioning, quickstart, and architecture/cost sections.
Reviewed changes
Copilot reviewed 5 out of 5 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
labs/2026-03-14-skill-forge.md |
New lab entry documenting the competition and findings (formatting issue in markdown tables). |
ideas.md |
New ideas/backlog document capturing follow-on concepts from the competition. |
README.md |
Expanded framing, quickstart, and architecture/cost explanations; updated command list and skill notes. |
.claude/skills/mobius-profile/scripts/show_profile.py |
New script that prints agent stats/matches and suggests challengers (has correctness issues in loss attribution and match filtering). |
.claude/skills/mobius-profile/SKILL.md |
New skill documentation instructing how to run the profile script and interpret results. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 0f454f1507
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
- Remove unused imports (json, row_to_dict) - Skip voided/undecided matches (winner_id=None) instead of counting as LOSS - Only count losses against the actual match winner, not all opponents - Key win/loss counters by slug instead of name to avoid collisions - Filter retired agents (elo_rating=0) and untested agents from challenger recommendations Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
labs/2026-03-14-skill-forge.md— First two-phase competition: 6 agents designed Claude Code skills, then 6 testers functionally validated them against the live DB. Full results, scores, and findings documented./mobius-profile— Competition winner. Single-script skill that shows agent deep-dives with match history, win/loss analysis, and challenger recommendations. Tested and working.ideas.md— Living backlog of concepts that emerged from the competition (forge skill, match replay, agent factory, self-audit).Key Finding
The generate → test two-phase pattern was more valuable than any individual skill. Design scores and functional scores diverged significantly — one skill scored 2nd in design but last in functional testing (shipped with crashing bugs). This pattern should become repeatable.
Relates to #5 (mobius improve self-diagnosing loop)
Test plan
/mobius-profiletested against live DB with 40 agents🤖 Generated with Claude Code