# AI Workflow Audit Skill

Last updated: 2026-06-10

The AI Workflow Audit is a free agent skill for founders and operators who need to decide which messy repetitive workflow is worth improving with AI. It runs five stages: opportunity discovery, discovery interview, workflow diagnosis, AI use-case selection, and solution design.

## Install Constraint

Install or copy the full directory: `skills/ai-workflow-audit/`.

Do not install only `SKILL.md`. The entry file links to templates, rubrics, and examples that are required for a complete audit.

Current source directory: https://github.com/ryan-hennebry/ai-operator/tree/main/skills/ai-workflow-audit

Raw entry file: https://raw.githubusercontent.com/ryan-hennebry/ai-operator/main/skills/ai-workflow-audit/SKILL.md

Publication gate: these URLs use the current working repo and may 404 until the repo is public and the final slug is ready.

## Key Files

- `SKILL.md` - entry point, operating rules, stages, and final report shape.
- `templates/discovery-interview.md` - stage 2 discovery interview.
- `templates/workflow-diagnosis.md` - stage 3 workflow map.
- `templates/ai-opportunity-scorecard.md` - stage 4 scoring and ranking.
- `templates/solution-design-spec.md` - stage 5 build-ready spec.
- `templates/audit-report.md` - final AI Workflow Audit Report structure.
- `rubrics/*.md` - pain, AI suitability, implementation readiness, and risk scoring.
- `examples/*.md` - illustrative example audits, not client deployment proof.

## Supported Agents

- Claude Code: install the full directory as a skill and invoke AI Workflow Audit.
- Codex: copy the full directory into a Codex skills location or keep the directory in the workspace and ask Codex to use it.
- Cursor: add the full directory to the workspace and ask the agent to run the skill from `SKILL.md`.
- Plain chat: use `SKILL.md` as the entry point and provide the linked templates, rubrics, and examples when needed.

## Output

The skill produces a 14-section AI Workflow Audit Report covering the current workflow, bottlenecks, AI opportunity scorecard, recommended first workflow, V1 system design, human review points, risks, eval criteria, implementation checklist, and suggested next step.

The report can also say not to build yet.