[recipes] Claudeception: continuous learning system that creates new skills

[jaredirish]

Summary

A meta-skill: skills that create other skills. Extracts reusable knowledge from work sessions and codifies it into structured skill files automatically.

When you discover something non-obvious (a debugging technique, an API quirk, a workaround), Claudeception evaluates whether it's worth preserving and creates a skill with trigger conditions, solution steps, and verification. Integrates with Open Brain to prevent duplicates and share knowledge across sessions.

Open Brain Integration

• Before creating: Searches search_thoughts to check if the knowledge already exists
• After creating: Captures the new skill via capture_thought so future sessions can find it

What Makes This Different

Every other OB1 recipe does a specific thing. This one creates new things from the act of working. It's the only recipe in the catalog where the output is more recipes/skills.

Files

• README.md - Full recipe with Prerequisites, Steps, Expected Outcome, Troubleshooting
• metadata.json - Recipe metadata
• claudeception.skill.md - The skill file with extraction process, quality gates, and OB1-relevant examples

Context

Fourth recipe contribution from @jaredirish. Part of the Open Brain Flywheel (capture-process-visualize loop). Reviewed via nate-advisor (Expert Lens pattern) before submission.

Previous contributions:

• [recipes] Panning for Gold - brain dump processor #21 Panning for Gold (merged)
• [recipes] Auto-Capture Protocol: the write side of the flywheel #42 Auto-Capture Protocol (open)
• [recipes] Infographic Generator: turn research docs into visual infographics #85 Infographic Generator (open)

Test Plan

• CI passes all 14 OB1 checks
• No credentials in any file
• README has Prerequisites, Steps, Expected Outcome, Troubleshooting
• Skill file includes Open Brain search-before-create and capture-after-create steps

[justfinethanku]

Code Review: Claudeception Recipe

This is a unique and valuable contribution to the OB1 ecosystem. The meta-skill concept (skills that create other skills) fills a genuine gap in continuous learning workflows. The quality of documentation and the Open Brain integration are both strong.

What's Good

Conceptual Strength:

• The meta-skill approach is genuinely innovative for the OB1 catalog. While other recipes add specific capabilities, this one creates a self-improving feedback loop.
• Open Brain integration at both ends (search-before-create, capture-after-create) demonstrates deep understanding of the OB1 philosophy.
• Quality criteria (reusable, non-trivial, specific, verified) show thoughtful consideration of over-extraction risks.

Documentation Quality:

• README has all required sections (Prerequisites, Steps, Expected Outcome, Troubleshooting) with clear, actionable content.
• The "Adapting for Other Tools" section makes this useful beyond just Claude Code users.
• Troubleshooting covers realistic edge cases (over-extraction, accumulation, missing OB integration).
• The credential tracker pattern is well-executed.

Technical Implementation:

• The skill file itself is comprehensive with a 7-step extraction process that includes web research and quality gates.
• Good balance between automation (triggers on non-obvious debugging) and manual control (/claudeception command).
• Anti-patterns section prevents common misuse.

metadata.json:

• All required fields present and valid.
• Appropriate difficulty level (intermediate).
• Realistic time estimate (10 minutes).
• Good tag coverage for discoverability.

Required Changes

None. This PR passes all 14 automated review checks and meets the contribution standards.

Nice-to-Haves (Optional Improvements)

1. Step verification checkpoints: While not required for recipes (they're strongly recommended for extensions), adding "✅ Done when:" lines after Steps 2, 3, and 5 would improve user confidence during setup.

2. Example skill in the repo: Consider adding a minimal example skill file (e.g., examples/sample-skill.md) showing what a completed extraction looks like. This would help users understand the target format before they create their first one. Not blocking, but would reduce friction.

3. Version discrepancy: The skill file shows version: 2.0.0 in its frontmatter (line 11), but metadata.json shows version: 1.0.0. Since this is the first release to OB1, 1.0.0 is correct for metadata.json. Consider updating the skill file to 1.0.0 or adding a note explaining that the skill file tracks its own versioning separately.

4. Cross-linking opportunity: The "Adapting for Other Tools" section mentions Cursor, Windsurf, and Codex but doesn't link to their documentation. Adding links would make it more actionable for users of those tools.

Safety & Standards Checks

✅ No credentials or secrets
✅ No dangerous SQL operations
✅ Folder structure correct (recipes/claudeception/)
✅ Required files present (README.md, metadata.json, claudeception.skill.md)
✅ PR title format correct: [recipes] ...
✅ No binary blobs
✅ README completeness verified
✅ All changes scoped to contribution folder
✅ Internal links resolve correctly (../../docs/01-getting-started.md exists)

Community Impact

This contribution aligns strongly with the Open Brain Flywheel concept (capture-process-visualize loop) mentioned in the PR description. It's the kind of meta-capability that makes the entire ecosystem more valuable over time.

The fact that this is the contributor's fourth PR shows consistent engagement with the project, and the mention of review via nate-advisor (Expert Lens pattern) demonstrates good quality control practices upstream.

---

Verdict: Ready to merge

This is a well-executed, thoroughly documented contribution that adds genuine value to the OB1 catalog. The only item worth addressing is the version discrepancy between the skill file and metadata.json, but even that is not blocking — it can be addressed in a follow-up if desired.

Great work, @jaredirish. Welcome addition to the recipes catalog.