| name | research-audit-graphics |
| description | Audit research topics for opportunities to add graphical representations (diagrams, graphs, schematics, tables, screenshots). Produces AUDIT directive comments for the refine phase. Arguments: optional topic path. |
| argument-hint | [topic-path] |
| disable-model-invocation | true |
| model | opus |
| allowed-tools | Read, Write, Glob, Grep, Bash(grep *), Bash(bash */skills/research-status/research-status.sh *), Edit, WebSearch, WebFetch |
Research Audit — Graphics
You are auditing research content for places that would benefit from graphical representations. You produce structured AUDIT comments that the refine phase will resolve.
Arguments: $ARGUMENTS
- First argument (optional): topic path relative to
research/content/ — a file or directory. Omit to audit all topics.
Prerequisites
- Derive each topic's status to identify which topics are in scope. Run
bash <skills-root>/research-status/research-status.sh research for the whole project (or --path <target> when a topic is targeted) and read the first whitespace-delimited field of each line. (<skills-root> is the .claude/skills/ directory the research skills are installed in — ~/.claude/skills for a global install, <project>/.claude/skills for a project install.)
- Topics with derived status
stub or inquiry are skipped (not yet ready for audit).
- Topics with derived status
draft, audited, or done are eligible.
- If a specific file is targeted and its derived status is
stub or inquiry, abort with an error.
- Read
research/CLAUDE.md for project conventions.
- Read all in-scope topic files.
- Read
research/DECISIONS.md for prior decisions.
- Read
research/glossary.md for term definitions.
Audit Operation: Graphics Opportunities
For each section, identify content that would be significantly clearer or more useful with a visual representation. Look for:
- Diagrams: processes, workflows, architectures, system interactions, state machines, decision trees — anything with components and relationships
- Graphs/Charts: quantitative comparisons, trends, distributions, performance characteristics, trade-off curves
- Tables: feature comparisons, option matrices, property summaries that are currently buried in prose
- Schematics: protocol flows, data formats, memory layouts, network topologies
- Screenshots: UI references, tool configurations, visual output that the text tries to describe
For each finding, assess whether the visual would materially improve comprehension — not every list needs a diagram. Prioritize cases where:
- The text describes spatial or relational structure that prose struggles to convey
- A comparison across 3+ items is written as sequential paragraphs
- A process with branching or parallelism is described linearly
- Quantitative relationships are described qualitatively ("much faster", "significantly larger")
For each finding:
Output
- Insert AUDIT comments directly into the topic files at the relevant locations (immediately after the content that would benefit from a visual).
- Update the
updated date in frontmatter for each modified file.
- Track audit progress in frontmatter: add or update an
audit field in each audited file's YAML frontmatter listing completed audit types — e.g. audit: [graphics]. If the field already exists, append graphics to the list (avoid duplicates).
- Present a summary to the user:
- Number of findings by graphic type and severity
- List of major findings (high-impact visuals that would substantially improve the content)
- Which files were modified
Note: the graphics audit is supplementary — appending graphics to the audit field does not affect the derived status (only the four core lenses advance a topic to audited). Do NOT treat this audit alone as advancing status.
Git
Do NOT commit. The user will review and use /commit when ready.
The expected commit message format: research(audit): <scope> graphics
Rules
- Do NOT modify content — only insert comments and update dates/frontmatter.
- Do NOT create the graphics — that is the refine phase's job.
- Severity guide:
major = content is genuinely hard to follow without a visual (complex process, multi-dimensional comparison). minor = a visual would be nice but the prose is adequate.
- If an AUDIT comment already exists at a location, do not duplicate it. Update the existing one if new information changes the assessment.
- Be selective. A handful of high-value graphics suggestions per topic is better than flagging every paragraph. Aim for visuals that earn their space.