get_watch_changes

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

Adds substantial context beyond the annotations: batches replay until acknowledged (even after restart), retries cannot confirm a newer batch, limit applies only to new batches, save=true writes a durable copy before acknowledgement and retries reuse it, source order is fair rather than globally sorted, and only the last 100 message IDs per source are checked. This is exactly the kind of non-obvious state, idempotency, and coverage-limit disclosure an agent needs before calling a non-read-only, non-idempotent poller.

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 opening purpose sentence is well front-loaded and every line carries information, but the remainder is an unstructured wall of prose mixing acknowledgement semantics, save behavior, ordering, and coverage limits. Some sentences ('Retrying that acknowledgement cannot confirm a newer batch') are dense enough to require re-reading, which hurts scannability for an agent.

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?

With no output schema and 4 undocumented parameters, the description does a lot of the missing work: it explains the replay/ack lifecycle, the saved_file return, issues/omissions and retry-delay signals, and the 100-ID revision window. What it never describes is the shape of a returned batch (what a batch contains besides message IDs), so the return contract is only partially covered.

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 0%, so the description must carry the load, and it does: name is for continuing incomplete passes, ack_batch_id acknowledges and polls in one call, limit applies only to new batches, and save=true writes a JSONL copy and returns saved_file. The gap is that save is said to 'require archive_directory,' which is not a parameter of this tool and is left unexplained, creating ambiguity about how to satisfy that prerequisite.

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?

States a specific verb and resource: 'Poll a named subscription without managing per-chat tokens,' plus it clarifies the batch/acknowledgement model. It differentiates itself from ack_watch_changes by naming it as the separate acknowledgement path, but it never distinguishes itself from the sibling get_changes, which sounds closely related.

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?

Gives concrete routing rules: use ack_batch_id to acknowledge and poll in one call, or use ack_watch_changes instead; set save=true when a permanent JSONL copy is wanted; 'continue incomplete passes by name.' It lacks an explicit statement of when NOT to use this tool versus get_changes or search_messages, so it stops short of full when/when-not guidance.

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