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

The annotations only say read-only and non-destructive, but the description discloses a non-obvious default: results are silently filtered to 'new since last read', which an agent could not infer from the schema or annotations. It also warns that mark_read=true has a lasting side effect, because the read cursor advances and the batch can no longer be retrieved, and it tells the agent the precondition for passing it. That is exactly the behavioral context annotations do not carry; the only tension is that this read-cursor mutation is not reflected in readOnlyHint, but the description itself is transparent about it.

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?

The text is front-loaded with what the tool returns before covering modes and the mark_read warning, and the ①② enumeration makes the two reply categories easy to parse. It is dense but every clause carries operational weight; only the bold emphasis and the run-on nature of the final sentence cost it a point.

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?

For a 3-parameter, no-required-argument retrieval tool, the description covers scope, default filtering, history mode, and the destructive-ish mark_read precondition, which is enough to call it correctly. The remaining gap is the shape of a returned reply item (author, post context, timestamp), which is undocumented since there is no output schema, but that is secondary to selection.

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 the baseline is 3, but the description adds operational meaning: since_hours is the escape hatch for historical lookback rather than a filter applied on top of the cursor, and mark_read is gated on user confirmation plus mode. It does not restate limit, which the schema already documents, so there is limited repetition. This is a modest but real improvement over the schema text.

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 defines the resource precisely and resolves the main ambiguity of the word 'replies' by enumerating two categories: comments/replies under posts I published, and replies directed at me inside someone else's thread. It also states an explicit exclusion ('自己回自己不算'), so an agent can tell exactly which objects are returned. This is far more specific than the title alone, and the scope wording implicitly distinguishes it from the per-post sibling get_post_replies.

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?

It gives clear when-to-use guidance for the two modes: the default returns only content newer than the last read cursor, while passing since_hours switches to historical lookback. It also gives strong when-not guidance for mark_read, restricting it to cases where the user has confirmed seeing the content or the agent has already relayed it. The only gap is that it never names the sibling tools (get_post_replies, get_post_detail) to route between them explicitly.

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