Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/data-wise/craft/plugin-auditnpx skills add Data-Wise/craft --skill plugin-auditgit clone --depth 1 https://github.com/Data-Wise/craftWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00169 | $0.02545 |
| Opus 5 | $0.00084 | $0.01273 |
| Sonnet 5 | $0.00034 | $0.00509 |
| Haiku 4.5 | $0.00017 | $0.00254 |
Grade A, and why
plugin-audit scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directorieslowAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
- Does not uninstall, disable, or edit any plugin, marketplace entry, or `~/.claude/settings.json` key. It reports; the user (or a follow-up command) acts. Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Plugin Audit
Finds duplicate or colliding plugins in the current Claude Code installation — the class of finding that a one-off manual sweep is prone to miss. Grew out of a 2026-07-01 session in which a manual 60-plugin review (2026-06-29) had missed workflow@local-plugins, a standalone plugin whose commands/workflow/* fully duplicated craft's own commands/workflow/* namespace (1131 legacy lines vs. craft's 112-line thin shim, same command names). See docs/specs/SPEC-token-efficiency-and-context-tooling-2026-07-01.md Part D #3.
Why this matters
Two plugins registering the same command or skill name — or near-duplicate content under different namespaces (workflow:brainstorm vs. craft:brainstorm) — is both a token-cost problem (duplicate content loads twice) and a correctness problem (routing ambiguity: which one fires?). Manual plugin audits are infrequent and don't scale past a handful of plugins; this skill makes the surface-diff repeatable and cheap enough to run any time plugin state changes.
What this skill does NOT do
- Does not uninstall, disable, or edit any plugin, marketplace entry, or
~/.claude/settings.jsonkey. It reports; the user (or a follow-up command) acts. - Does not evaluate whether a plugin's content quality is good — only whether its surface (command/skill names) collides with another installed plugin's surface.
- Does not require network access — everything it reads is local (
~/.claude/settings.json,~/.claude/plugins/, each plugin's own repo/cache directory).
Inputs
~/.claude/settings.json—enabledPluginsobject. Keys are<plugin-name>@<marketplace-name>; only plugins with a truthy value are actually active.~/.claude/plugins/installed_plugins.json(if present) — richer per-plugin metadata (source path, marketplace, version) thansettings.jsonalone provides. Treat as supplementary, not required — some installations only havesettings.json. Schema:{"version": 2, "plugins": {"<name>@<marketplace>": [{"installPath": "...", "version": "...", ...}]}}—.pluginsis an object keyed by<name>@<marketplace>, each value an ARRAY of install records (usually one). The on-disk path field isinstallPath, notpath.- Each enabled plugin's own directory — resolve via
installed_plugins.json's recorded path, or~/.claude/plugins/marketplaces/<marketplace>/<plugin>//~/.claude/plugins/repos/...(layout varies by install method: local marketplace vs. GitHub marketplace vs. Desktop). Within that directory, the actual surface is:commands/**/*.md— each file's path (minus the.md) becomes a command name, namespaced as<plugin>:<relative-path-without-ext>(e.g.commands/brainstorm.md→<plugin>:workflow:brainstorm).skills/**/SKILL.md— eachSKILL.md's parent directory name is the skill name, namespaced the same way.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 145 lines · 169 tokens per session scan A 151a243601c8
plugin-audit is a skill published in the GitHub repository Data-Wise/craft (4 stars, last pushed 17d ago), licensed MIT. It adds 169 tokens to every session and 2,545 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
architecture-review
Conducts a comprehensive multi-perspective architecture review using ALL architecture team members. Use when the user requests "Start architecture review", "Full architecture review", "Review architecture for version X.Y.Z", "Conduct comprehensive review", or when they want assessment from multiple perspectives. Do…
architecture-status
Reports on the health and state of architecture documentation (counts of ADRs, reviews, activity levels, documentation gaps). Use when the user asks "What's our architecture status?", "Show architecture documentation", "How many ADRs do we have?", "What decisions are documented?", "Architecture health check", or wants…
pragmatic-guard
Enables and configures Pragmatic Guard Mode (YAGNI Enforcement) to prevent over-engineering. Use when the user requests "Enable pragmatic mode", "Turn on YAGNI enforcement", "Activate simplicity guard", "Challenge complexity", or similar phrases.
specialist-review
Conducts a focused review from ONE specific specialist's perspective (e.g., Security Specialist, Performance Expert). Use when the user requests "Ask [specialist role] to review [target]", "Get [specialist]'s opinion on [topic]", "Have [role] review [code/component]", or when they want deep expertise in ONE specific…
create-adr
Creates a NEW Architectural Decision Record (ADR) documenting a specific architectural decision. Use when the user requests "Create ADR for [topic]", "Document decision about [topic]", "Write ADR for [choice]", or when documenting technology choices, patterns, or architectural approaches. Do NOT use for reviews (use…
list-members
Displays the roster of architecture team members with their specialties and expertise areas. Use when the user asks "Who's on the architecture team?", "List architecture members", "Show me the architects", "What specialists are available?", "Who can I ask for reviews?", or wants to discover available experts. Do NOT…