Diagnoses Pixie tool failures and proposes validated fixes using kill file and run logs.
npx skills add https://github.com/AlexKapadia/Pixie --skill debug-tool
Please help me install this Agent Skill. Skill: debug-tool Source: https://github.com/AlexKapadia/Pixie/tree/main/.claude/skills/debug-tool Command: npx skills add https://github.com/AlexKapadia/Pixie --skill debug-tool
Diagnoses and fixes ONE Pixie tool that fails validation, crashes, hangs, or returns wrong output – validator-first. Use when the user reports a broken tool, a tool that crashes, or outputs incorrect results. Do NOT use if Pixie itself is broken (pixie-doctor), for style (lint-tool), or just logs (view-logs).
Diagnose a failing tool by starting with the validator, then follow its guided steps to identify, read the kill file, propose a fix, and re-run validation.
High-intent search queries and answers about installing and using this skill.
To diagnose a Pixie tool that fails validation, use a validator-first debugging workflow that runs the Pixie validator to locate the failure, reads the kill file for documented fixes, and proposes concrete code or configuration edits to resolve the crash.
The best way to fix a Pixie tool returning incorrect output is to run the validator, consult the run logs and kill file entries for those specific symptoms, and apply the guided remediation steps before re-running validation to confirm the repair.
You should use a validator-first workflow to debug a misbehaving tool when a named tool fails validation, crashes, hangs, or produces incorrect output, ensuring you locate the precise failure point before attempting any code or configuration repairs.
To use a kill file for fixing a failing tool, read the .build/KILL_FILE.md entries related to the tool's symptoms, apply the documented fixes to your code or configuration, and re-run the validator to verify the tool passes.
No, you cannot use this debugger if Pixie itself is broken, as that requires pixie-doctor, nor is it meant for style issues requiring a lint-tool or simple log viewing, because it strictly diagnoses and fixes one misbehaving named tool.
To re-validate a tool after applying a proposed fix, execute the command uv run pixie validate, which checks if the structured edits successfully resolved the validation failure and confirms the tool now passes the runtime checks.
AlexKapadiaRe-run Pixie validation across all tools and summarize pass/warn/fail results.
AlexKapadiaConvert ML model specifications into Pixie tool scaffolds with training and inference workflows.
AlexKapadiaGenerate a read-only markdown inspection report for a single Pixie tool.
AlexKapadiaLint a single Pixie tool for metadata, imports, secrets, and log hygiene.
