From workflow-optimizer
Bucket failures from a measurement run into actionable categories (TIMEOUT, AUTH, UI_CHANGE, etc). Standalone — does not invoke other skills.
How this skill is triggered — by the user, by Claude, or both
Slash command
/workflow-optimizer:categorizeThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Path to a baseline or measurement file containing run results with error messages.
Path to a baseline or measurement file containing run results with error messages.
For each failed run, match error text against these patterns (first match wins):
| Category | Patterns | Fixable? |
|---|---|---|
| AUTH | logged out, sign in, expired session, unauthorized | No — manual re-login |
| RATE_LIMIT | 429, rate limit, too many requests | No — wait required |
| TIMEOUT | timeout, timed out, deadline, turn limit exceeded | Yes |
| UI_CHANGE | selector not found, element not found, no such element | Yes |
| TIMING | not ready, loading, spinner, pending | Yes |
| CONTENT | text empty, failed to type, paste fail | Yes |
| BROWSER | browser crash, chromium, target closed, page closed | Partial |
| UNKNOWN | (no pattern matched) | Review needed |
For each failed run:
error + output into a single search text{ category, error, fixture_id }Failures (3/5 runs):
TIMEOUT 2 (fixable)
AUTH 1 (not fixable)
Fixable: 2/3 (67%)
TIMEOUT (2):
1. [f1] Timed out after 300s
2. [f2] Turn limit exceeded at 30 turns
AUTH (1):
1. [f3] Session expired, login required
| Condition | Recommendation |
|---|---|
| All AUTH or RATE_LIMIT | SKIP — not fixable by code changes |
| Majority fixable categories | FIX — proceed to optimization |
| All UNKNOWN | REVIEW — examine raw output manually |
| Mix of fixable + unfixable | FIX — address fixable, note unfixable |
npx claudepluginhub yihan2099/workflow-optimizer --plugin workflow-optimizerClassifies errors into 7 taxonomy categories (syntax, logic, design, performance, etc.) and routes to deterministic resolution strategies. Use when root cause is unclear or bug fix attempts fail.
Analyzes failing test cases to classify failures as test bugs, implementation bugs, or ambiguous. Uses investigative process to verify expectations against code and docs before recommending fixes.
Guides analysis of LLM pipeline traces to identify, categorize, and prioritize failure modes. Use for new eval projects, pipeline changes, metric drops, or incidents.