jbaruch/nanoclaw-untrusted
Security rules for untrusted NanoClaw groups. Credential protection, internal file protection, social engineering defenses.
80
100%
Does it follow best practices?
Impact
—
No eval scenarios have been run
Passed
No known issues
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a well-crafted skill description that clearly articulates specific capabilities, provides explicit trigger conditions via a 'Use when...' clause, and targets a distinct niche around permission and content boundary management in group settings. It uses proper third-person voice throughout and includes natural language trigger terms that users would realistically use.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | The description lists multiple specific concrete actions: 'Lists permitted and prohibited actions', 'blocks disallowed content types', and 'responds to permission queries'. These are clear, actionable capabilities. | 3 / 3 |
Completeness | Clearly answers both 'what' (lists permitted/prohibited actions, blocks disallowed content, responds to permission queries) and 'when' with an explicit 'Use when...' clause covering multiple trigger scenarios. | 3 / 3 |
Trigger Term Quality | Includes natural trigger terms users would say: 'rules', 'permissions', 'boundaries', 'what you are allowed to do here', 'public channel', 'group chat', 'joining a new group'. These cover a good range of natural phrasings. | 3 / 3 |
Distinctiveness Conflict Risk | The description carves out a clear niche around group/public channel permission management and content moderation boundaries. The specific context of 'shared or public group settings' and 'untrusted group chat environment' makes it highly distinctive and unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is an excellent, lean behavioral skill that clearly defines boundaries for operating in untrusted group environments. It provides specific, actionable constraints with concrete examples of what to say and what not to disclose, all in a highly scannable format. The content respects Claude's intelligence while adding genuinely useful guardrails that Claude wouldn't know by default.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every line earns its place. No unnecessary explanations of what a group chat is or how AI assistants work. The guidelines are terse, direct, and assume Claude's intelligence. | 3 / 3 |
Actionability | Despite being an instruction-only skill (no code needed), the guidance is highly specific and concrete: exact phrases to use ('I can't do that in this chat'), specific behaviors to avoid (don't explain container isolation, don't attempt writes), and clear decision rules (follow owner's lead, decline when in doubt). | 3 / 3 |
Workflow Clarity | This is a simple behavioral skill with no multi-step process. The single-task guidance is unambiguous: a clear list of permitted actions, followed by prioritized behavioral rules with explicit decision criteria for ambiguous situations. | 3 / 3 |
Progressive Disclosure | For a simple skill under 50 lines with no need for external references, the content is well-organized into two clear sections (what you can do, behavioral guidelines) with bolded topic sentences for easy scanning. No external files are needed. | 3 / 3 |
Total | 12 / 12 Passed |
Validation
100%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 11 / 11 Passed
Validation for skill structure
No warnings or errors.
Reviewed
Table of Contents