feature-guide — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited feature-guide (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Generates a structured feature card for any Claude capability — given a feature name or a description of what the user wants to accomplish. Covers: what it is, where it's available, how to use it, limitations, current status, and an applicability verdict.
Russian: «что умеет Claude», «есть ли у Claude», «умеет ли Claude», «поддерживает ли Claude», «как включить фичу», «расскажи о фиче», «scout», «какие возможности у Claude», «как использовать», «feature scout» English: "what can Claude do", "does Claude have", "does Claude support", "how to enable", "Claude feature", "scout", "can Claude", "what features", "how to use", "feature scout", "Claude capabilities"
Determine the language of the user's request:
Required: a feature name OR a description of what the user wants to accomplish (free text, 1–2 sentences is enough)
Examples:
One feature card per request, in structured format.
Classify the incoming request:
| Type | Signal | Action |
|---|---|---|
| Direct feature name | User names a specific feature: "Projects", "Memory", "Voice" | Proceed to Step 2 |
| Need description | User describes a goal: "I want Claude to remember things" | Identify the closest feature, ask one clarifying question |
| Ambiguous | Multiple features match | Offer up to 3 options, ask user to choose |
| Too general | "What can Claude do?" with no further context | Ask one question: "What specifically do you want to accomplish with Claude?" |
| AI comparison | Request mentions another AI product: ChatGPT, Gemini, Copilot, etc. | Decline: "This skill covers Claude's features only. Want to explore [feature from request]?" |
Do not ask more than one clarifying question. When context is minimal, act on the most probable interpretation.
Run a web search to get the feature's current status. Priority sources:
site:docs.anthropic.com [feature name]site:support.anthropic.com [feature name]Claude [feature name] 2025 OR 2026Goal: current status (available / beta / Pro+ only / announced / not released), platforms, and activation steps.
If web search is unavailable — continue based on training data and mark the card: "Based on training data — status may be outdated."
Determine each parameter from search results:
| Parameter | How to determine |
|---|---|
| Name | Official name from Anthropic documentation |
| Description | 1–2 sentences: what it does, why it matters to the user |
| Platforms | Claude.ai / Projects / Claude Code / API / Cowork / mobile app |
| Plans | Free / Pro / Team / Enterprise / all plans |
| How to activate | Specific steps or trigger phrase |
| Limitations | What the feature does NOT do, technical limits |
| Status | Available / Beta / Pro+ only / Announced / Not released |
| Verdict | Applicability to the user's specific goal |
If a parameter is unknown — write "No data". Do not fabricate information.
Output the card strictly in the following format:
What it is: [1–2 sentences]
Where available: [platforms] — [plans]
How to use: [specific steps or trigger phrase]
Limitations:
Status: [Available / Beta / Pro+ only / Announced / Not released]
Verdict: [Fits / Partially fits / Does not fit] — [1 sentence with reasoning]
After the card, add one line: "Explore related features: [name 1], [name 2]?" — only if related features are clearly relevant to the request.
| Situation | Action |
|---|---|
| Feature does not exist | Do not generate a card. Output: "Feature [name] not found in Claude documentation. Related features: [name 1], [name 2] — explore?" |
| Feature announced but not released | Status: "Announced", include date if known |
| Feature is Claude Code only, non-technical audience | Add to Limitations: "Requires technical setup" |
| Different versions across platforms | Note differences under "Where available" |
| Request covers multiple features at once | Generate card for the first feature, offer to continue with the rest |
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.