Claude Code plugins

Two plugins of Claude Code skills: audit (review workflows) and workflow (project automation). Most skills are invoked via slash command rather than auto-triggered; see each plugin's README for exact invocations. Background on the underlying mechanism: Claude Skills documentation.

Install

Add the marketplace, then install plugins individually:

claude plugin marketplace add hebstr/claude-code-plugins
claude plugin install audit@hebstr
claude plugin install workflow@hebstr

Plugins

audit

Code, skill, and MCP review workflows: orchestrated walkthroughs, full-project audits, adversarial critique, and circularity-aware cross-model review.

Skill	Purpose
walkthrough	Interactive, point-by-point walkthrough of any review report: re-evaluates findings, proposes fixes, checks impacted files for regressions
sweep	Full-coverage project review with parallel specialist agents (architecture, quality, tests, docs), consolidated into a single deduplicated report
blindspot	Circularity-aware orchestrator that detects when a reviewer shares its target's codebase or model family, then injects cross-model judging via OpenRouter
skill-adversary	Adversarial critic for Claude Code skills: finds trigger edge cases, instruction ambiguities, cross-file coherence issues, and gaps
mcp-adversary	Adversarial critic for MCP servers: finds inter-tool discrimination issues, schema anti-patterns, semantic drift, and undocumented workflow dependencies

workflow

Project workflow automation: file consistency sweeps, cross-repo synchronization, prose editing, session handoffs, source-backed recommendations, and documentation layout audits.

Skill	Purpose
sync	Scan all files, identify ones that are stale relative to recent changes, and update them. Always runs a cross-repo semantic consistency scan with parallel agents
write	Strip AI writing patterns from prose and rewrite it to sound human. Routes to a French or English reference based on the text being edited. Includes a bilingual review mode (FR↔EN parity, typography, faux amis)
continue	Flush durable facts to memory, update .claude/PLAN.md, and print a minimal continuation prompt. No handoff document is written; PLAN.md and memory are the authoritative stores
reco	Deep-mode recommendation backed by external sources. Parallel agents research official docs (WebFetch) and community practice (WebSearch), then synthesize a structured recommendation with verified citations
doc-structure	Audit project documentation layout (CLAUDE.md vs README.md), propose verbatim migrations of misplaced prose, and update the CLAUDE.md index. Adapted from solatis/claude-config:doc-sync

Requirements

Required for both plugins:

• Claude Code with plugin marketplaces enabled
• git (review targets and workflow:sync operate on git state)

Optional, per feature. Skills degrade gracefully when a dependency is absent: the missing feature is reported, and the rest of the skill continues.

Dependency	Used by	Unlocks
OPENROUTER_API_KEY env var (+ jq)	audit:blindspot, audit:walkthrough	Cross-model judging and L2 cross-provider validation via OpenRouter
Ouroboros plugin	audit:walkthrough	Consensus QA, evaluate + drift checks, lateral-think rescue
critical-code-reviewer (posit-dev/skills)	audit:sweep	Specialist code-review agent; sweep falls back to inline review when absent
r-lib skills testing-r-packages, r-package-development, cran-extrachecks (posit-dev/skills)	audit:sweep on R projects	R-specific test, package, and CRAN-readiness audits
gh CLI	audit:walkthrough	PR body as evaluate context (falls back to latest commit message)

The workflow plugin has no external dependencies beyond Claude Code and git. See audit/README.md for per-skill detail.

License

MIT