3.9 KiB

Raw Blame History

Workflow: Audit Workflow

Evaluate workflow documents for quality, compliance, and completeness against workflow-architect standards.

Who should do this work

Haiku subagent: Research phase (setup, context loading, structure audit)
Sonnet subagent: Analysis phase (evaluation, quality assessment)
Opus subagent: Synthesis phase (classification, verdict, report)
Haiku subagent: Cleanup phase

When to use

Evaluating workflow for quality and compliance
Checking workflow-architect standards
Reviewing structure, clarity, completeness

Inputs required

workflow_path - Path to workflow file (e.g., plugin/skills/my-skill/workflows/my-workflow.md)

Output

.claude/plugin-data/workflow-architect/audits/<workflow-name>-audit.md (gitignored)
Summary with verdict (Ready / Not Ready)
Delete report after user acts on findings

Phase 1: Research

Subagent: haiku. Read: ${CLAUDE_PLUGIN_ROOT}/references/layout/workflow-structure.md

Setup:

Verify workflow_path exists, is .md file
Verify in workflows/ directory
Create scratch workspace
Extract workflow name

Context loading:

Check for invariants.md in workflow's plugin root
Check for .decisions/ directory
Check for examples/golden/
Record: invariants list, prior decisions count/topics, golden examples list (or "none" for each)

Structure audit:

Has "When to use"
Has "Inputs required"
Has "Output"
Has "Subtasks" or "Phase" sections with clear steps
Each subtask/phase: "Who" (main/subagent)
Each subtask/phase: "Must read" if applicable
Each subtask/phase: clear "Do" instructions

Read workflow file:

Full workflow content
Note all referenced files

Write: scratch/research-findings.md

Phase 2: Analysis

Subagent: sonnet. Read: scratch/research-findings.md, ${CLAUDE_PLUGIN_ROOT}/references/layout/workflow-structure.md

Ownership clarity evaluation:

Each subtask/phase clearly assigns ownership (main vs subagent)
Subagent tasks reference model if specified
Main tasks don't delegate subagent work
No ambiguous ownership

Completeness evaluation:

Clear purpose and trigger conditions
Inputs well-defined
Output format specified
All subtasks/phases have necessary context
Verification steps if modifies files
No missing steps in workflow flow

Quality assessment:

Logical step ordering
Appropriate subagent model assignments
File artifacts properly passed between phases
No redundant steps

Write: scratch/analysis-findings.md

Phase 3: Synthesis

Subagent: opus. Read: scratch/analysis-findings.md, ${CLAUDE_PLUGIN_ROOT}/references/tool-patterns/severity-guide.md, references/tool-patterns/catalog.md

Classify findings:

Classify by severity (significant/minor/polish)
Verdict: Ready (0 significant) or Not Ready (1+ significant)

Severity for workflows:

Significant: Missing sections, ambiguous ownership, incomplete flow, missing verification
Minor: Unclear wording, missing optional context, suboptimal structure
Polish: Style, formatting

Impact analysis per recommendation:

What could degrade?
Invariant conflict?
Golden example impact?

If contradicts invariant or affects golden example: note and require user approval.

Write: .claude/plugin-data/workflow-architect/audits/<workflow-name>-audit.md

Present report to user:

Walk through significant findings first
Offer to fix or let user handle
After user confirms: invoke cleanup

Cleanup (haiku):

Delete scratch workspace
Delete audit report
Record decisions if needed:
- Template: references/tool-patterns/templates/decision-record-template.md
- Write: <plugin_path>/.decisions/YYYY-MM-DD-<topic>.md
- When: non-standard structures, exceptions to best practices, complex delegation patterns
- Prompt: "Were any significant decisions made? (structure exceptions, pattern decisions, best practice deviations)"

3.9 KiB Raw Blame History