Use after a complex task, failure, or when reviewing what was learned. Teaches how to write growth logs that extract reusable patterns — not diary entries.
Growth Log is a methodology skill that teaches how to write learning entries that compound across sessions instead of shallow diary notes. It targets a common failure: recording 'fixed a bug in X', which captures an event but not the reusable pattern. The skill reframes each entry around extracting the underlying pattern so the same class of problem is recognized and handled faster next time. It is note-system agnostic, working with Markdown files, Notion, Obsidian, or plain text, with generic templates to adapt.
The core is three rules: failures are more instructive than achievements and deserve richer entries; the Bole Principle says to search existing notes and merge same-root-cause entries rather than duplicating; and every entry must be transferable, containing an explicit 'next time I see [signal], I will [action]' sentence. It provides a five-step pattern-extraction process (state, ask why iteratively, generalize, formulate the action, name the signal), a reusable entry template with Context / Root Cause / Pattern / Related sections, four entry types (Failure, Methodology, Pattern Discovery, Capability Change), a quality checklist, and anti-patterns to avoid. Guidance on scope keeps entries to roughly 4-8 sentences.
It is aimed at developers and AI-agent operators who want deliberate, searchable learning capture, and it pairs naturally with the delivery-gate Stop hook: the hook enforces that a learning file was touched, while this skill ensures the content is a useful pattern rather than an empty timestamp. It is purely instructional documentation with no code, tools, or system access.
After a complex, multi-file task, after a failure or rework, or when reviewing what you learned over a period. Skip trivial changes like typo fixes; the threshold is whether the task involved debugging, redoing, rollback, or a non-obvious decision.
It names the pattern in its title, distinguishes root cause from symptom, and includes a transferable 'next time I see [signal], I will [action]' sentence with a specific recognizable signal.
No. It works with any note system — Markdown files, Notion, Obsidian, or plain text. The templates are generic; you pick one storage convention and stick to it for searchability.
Typically 4-8 sentences, one per distinct root cause. Shorter suggests you haven't reached the pattern; much longer means you're narrating events instead of extracting insight.
The delivery-gate Stop hook checks via filesystem timestamps that a learning file was modified today; this skill teaches what to write so that file contains a useful pattern rather than an empty entry.
Quick Setup:
.claude/skills/Repository
affaan-m/ecc