Review a generated report for narrative coherence, internal contradictions, and unsupported assumptions — with no access to source data or analysis context. Delegated by the report generation workflow as an optional quality gate. <example> Context: Workflow delegates coherence review after report assembly and validation user: "Review the report at /path/to/report.html for coherence issues" assistant: "I'll read the report with fresh eyes and check for narrative coherence, contradictions, and missing context." <commentary> The coherence-reviewer receives ONLY the report file path. It has no access to analysis data, source code, or git history — simulating a first-time reader's experience. </commentary> </example>

Analyze functionality, architecture, dependencies, and quality of agent plugin components (Claude Code). Delegated by the agent-extension-visualizing skill. <example> Context: Skill delegates feature analysis with metadata and file paths user: "Analyze features for plugin at ./plugins/my-plugin with components: [SKILL] my-skill, [AGENT] my-agent" assistant: "I'll analyze functionality, architecture, dependencies, and quality of each component." <commentary> The agent-extension-visualizing skill provides metadata and file paths. This agent reads the actual files and performs feature/architecture analysis. </commentary> </example>

Analyze security posture, permission models, and risk levels of agent plugin components (Claude Code). Delegated by the agent-extension-visualizing skill. <example> Context: Skill delegates security analysis with metadata and file paths user: "Analyze security for plugin at ./plugins/my-plugin with components: [SKILL] my-skill, [AGENT] my-agent" assistant: "I'll audit the permission model, tool scope, hook scripts, and MCP trust boundaries." <commentary> The agent-extension-visualizing skill provides metadata and file paths. This agent reads the actual files and performs security analysis. </commentary> </example>

Generate self-contained HTML reports from structured analysis data with dynamic section structure. Report generator for all vision-powers visual report skills (agent-extension-visualizing, diff-visual, plan-visual, project-recap). <example> Context: Skill delegates HTML report generation with analysis results, section structure content, and design system content user: "Generate HTML report to ${CLAUDE_PLUGIN_DATA}/reports/my-diff-report.html" assistant: "I'll generate a self-contained HTML report following the provided section structure and design system." <commentary> The orchestrator skill provides pre-analyzed data and reference file contents inline (section structure, font system, anti-slop rules). This agent starts writing sections immediately without a read turn. </commentary> </example>

Analyze agent extensions and generate self-contained HTML wiki reports with security audit, architecture diagrams, and plugin profiles. Currently supports Claude Code plugins. Use when asked to analyze, audit, inspect, review, document, or wiki a plugin or extension — including phrases like "what does this plugin do", "tell me about this extension", "break down this plugin", or "generate a report for this plugin". Also triggers on GitHub plugin URLs or local plugin paths. Default output is an interactive HTML report; use --format md for inline markdown.

Visualize git diffs as interactive HTML reports with architecture diagrams, KPI dashboards, code review cards, and side-by-side comparisons. Use when asked to visualize, review, explain, or summarize a diff, branch, commit, PR, or set of changes — including phrases like "what changed", "show me the changes", "review this PR visually", or "make a visual diff report". Accepts branch names, commit hashes, HEAD, PR numbers, or commit ranges.

Verify factual accuracy of a document against the actual codebase and git history. Extracts verifiable claims, checks each against source, corrects inaccuracies in place, and adds a verification summary. Use when asked to verify, fact-check, validate, or audit claims in a report, plan, or document — including phrases like "is this accurate", "double-check this report", "verify the numbers", or "are these claims correct". Accepts a file path or auto-detects the most recent HTML report.

Review implementation plans as interactive HTML reports with architecture diagrams, blast radius analysis, risk assessment, and gap detection. Use when asked to review, visualize, evaluate, assess, or critique an implementation plan — including phrases like "check my plan", "is this plan feasible", "review the implementation plan", or "what could go wrong with this plan". Accepts plan file paths or reads from the current Claude Code plan.

Generate a visual project recap — rebuild mental model of a project's current state, recent activity, key decisions, and cognitive debt hotspots. Use when asked to recap, summarize, snapshot, or catch up on a project's status, progress, or recent activity — including phrases like "what happened recently", "catch me up", "status update", "what's been going on", or "give me the big picture". Accepts a time window (2w, 30d, 3m).

Manage and refine vision-powers reports: list, open, delete, search, and refine sections of generated HTML reports. Use when asked to list reports, open a previous report, delete old reports, find a specific report, or fix/update a section of an existing report — including phrases like "show my reports", "open the last report", "clean up old reports", "find a report about X", "fix section 3", or "update report".

1 hook across 1 event