From canon
This skill should be used before suggesting an implementation approach for a problem that resembles past work — when the user asks "how should I fix X", "what's the best way to do Y", "let's tackle Z", or whenever a non-trivial approach is about to be proposed. Also triggers on "log this failure", "add to ERRORS.md", or "this didn't work, remember it for next time".
How this skill is triggered — by the user, by Claude, or both
Slash command
/canon:errors-checkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Two-way skill for the failure log:
Two-way skill for the failure log:
ERRORS.md for prior failures on similar problems and surface the rejected approaches.This is the "stop suggesting things we already tried" mechanism. It is what keeps Claude from re-proposing Drizzle three months after the team picked Prisma for a specific reason.
Read mode — trigger silently before proposing implementation for:
Don't ask the user — just check, then mention what was found if anything matches.
Write mode — trigger on phrases like:
ERRORS.md at the project root. If absent, skip silently — don't pester the user.ERRORS.md at the project root. If absent, suggest creating it (or running canon-init).## YYYY-MM-DD: <one-line description of the problem>
- **What didn't work:** <the failed approach>
- **Why it failed:** <root cause, as best understood>
- **What worked instead:** <the actual fix>
- **Note for next time:** <one-liner future-Claude should remember>
- **Files / areas affected:** <if applicable>
Per Karpathy's original framing: log an entry only when an approach took more than two attempts to get right. One-shot fixes don't belong in ERRORS.md — they belong in the implementation. The signal-to-noise ratio of ERRORS.md is what makes it useful; flooding it with trivial misfires defeats the purpose.
npx claudepluginhub orthogon-ai-labs/canonGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.