Audit skills, commands, agents and CLAUDE.md

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare readOnlyHint=true and openWorldHint=false, yet the description still earns credit by adding that it 'launches nothing and sends nothing' — meaningful because an agent may otherwise fear that scanning hook configs or bundled scripts executes or exfiltrates them. The enumerated detection categories (invisible text, credential-file directives, download-and-run, pre-approved unrestricted shell) further describe behavioral scope beyond what annotations convey, though it says nothing about scan cost or how it handles symlinks/large trees.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with what is scanned and where, before the threat list, which is the right order for an agent deciding whether to call it. The second sentence's long enumeration of a dozen threat categories is dense but each item denotes a distinct, non-redundant detection class, so it mostly earns its length.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

No output schema exists, so the description must imply results; the flag list effectively enumerates finding categories, which is adequate for an agent to expect structured findings. It stops short of describing result shape (file paths, severity, remediation) or what happens on a clean scan, and there is no pointer to follow-up siblings like apply_fixes or generate_policy.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so baseline is 3, but the description adds scope semantics the schema lacks: scanning spans the user level, the project, and every enabled plugin by default, which explains what project_only actually narrows away. It doesn't add format or path-behavior detail for project_dir beyond the schema's default note.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a specific verb and enumerates the exact resource class scanned (CLAUDE.md, skills and bundled scripts, slash commands, subagents, rules, plugin hook configs) plus the three sources (user, project, enabled plugins). The phrase 'besides MCP tools' explicitly carves out the scope handled by sibling tools like audit_mcp_config, so an agent can discriminate without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Does the description explain when to use this tool, when not to, or what alternatives exist?

Usage is implied rather than stated: the catalog of scanned sources and the threat list signal 'audit before trusting agent config,' and the project_only param mentions 'repositories you are about to trust,' but that guidance lives in the schema, not the description. The description never names an alternative tool (audit_mcp_config, audit_server_tools) or an explicit precondition such as 'run before installing a plugin.'

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.