Menu
Debug a workflow YAML — validate, trace execution, simulate state flow, and catch issues before runtime.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Deep analysis of current changes before committing. Goes beyond lint — catches logic bugs, architectural violations, security, and performance issues.
Create a pull request from the current branch, matching the repo's PR conventions (What/Why/Testing for features, Summary/Test-plan for small PRs).
Review staged or unstaged changes before committing. Catches bugs, style drift, and missing coverage.
Cut a release — promote Unreleased → versioned in CHANGELOG, bump version, commit. CI auto-tags and publishes from the version-bump commit.
Create a release-tracking issue that organizes open issues into a phased roadmap. Pass a target version (e.g., `0.5.0`) to scope the plan.
Run the full quality gate — format, lint, tests, types. Use after changes to verify nothing is broken.
| name | debug-workflow |
| description | Debug a workflow YAML — validate, trace execution, simulate state flow, and catch issues before runtime. |
Debug the workflow at the path given in $ARGUMENTS (or ask for one).
uv run agentloom validate $ARGUMENTSuv run agentloom visualize $ARGUMENTSRead the YAML and trace execution mentally:
output: field. Does each step's prompt: reference variables that will exist at that point? Walk through the state dict step by step.state.x is a dict, not a string)?tool_name a registered tool? Are tool_args valid for that tool's schema? Read tools/builtins.py to check.budget_usd is set, estimate cost per step (model + estimated tokens). Will it exceed?For each step in execution order, write out:
Step: <id> (<type>)
Input state: {key: expected_value, ...}
Expected output: <what this step produces>
State after: {key: new_value, ...}
Risk: <what could go wrong>