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

"Read-only" is consistent with readOnlyHint=true, and the description adds context the annotations do not: paged results via offset/nextOffset, and the critical operational note that allowed directories are "(none configured)", which tells an agent calls may fail or be scoped to nothing. The depth/slowness tradeoff is stated but partly duplicates the deep param's schema text.

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 the primary output, then the deep mode, then the safety/paging constraints. Dense but every clause carries information; the parenthetical "(candidate formats, sizes, quality target, output path)" is the only slightly verbose stretch.

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 12 parameters, nested objects and no output schema, the description does the minimum: it names the reported fields and the plan contents, but never describes the result envelope (per-file entries, nextOffset, error behavior for disallowed paths). The biggest gap is failing to say that policy/target/widths/resize/crop only affect the displayed plan, leaving an agent to guess.

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 75%, so the baseline is 3. The description explains deep, paths and offset/nextOffset paging, but ignores the nine output-shaping parameters (crop, resize, widths, target, policy, metadata, priority, limit) — an agent reading only the description would not know these merely annotate the deep plan rather than change the scan.

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 (scan/report) and resource (image files, folders, globs) plus exactly what is reported: format, size, dimensions, transparency, animation. It also distinguishes its two modes (shallow vs deep) and names the sibling whose plan it previews, so an agent can separate it from optimize_images 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?

Gives a clear trigger for the expensive mode ("With deep: true, also classify...") and implicitly routes the agent to optimize_images when it wants to act on the reported plan. It does not state when to prefer compare_images or when not to use this tool, so it stops short of explicit alternatives.

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

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

Annotations already declare idempotentHint=true, destructiveHint=false, readOnlyHint=false, and the description reinforces them non-contradictorily while adding real value: partially written images persist and are recorded in the manifest so a re-run skips them. That partial-completion/resume semantic is not derivable from any structured field. It stops short of stating latency, error behavior, or what a cancelled-by-race job returns.

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?

Two tight sentences with zero waste: the action is front-loaded and the consequences follow. Every clause carries information an agent needs.

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 single-parameter mutation tool with no output schema, annotations plus the partial-completion note cover nearly everything an agent needs. The only genuine gap is the unexplained jobId and its provenance.

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?

There is one required parameter, jobId, with 0% schema description coverage, so the schema supplies no meaning at all. The description never mentions jobId, where to obtain it (presumably from optimize_images or get_job), or what format it takes, so it fails to compensate for the coverage gap.

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 ('Stop') and a specific resource (an optimize_images job), tying it directly to the sibling optimize_images. An agent can immediately distinguish this job-control tool from analyze_images, get_job, and compare_images.

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 by 'Stop an optimize_images job', but there is no explicit when-to-use guidance, no mention of get_job for checking status first, and no statement of when cancellation is inappropriate (e.g. already-finished jobs).

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

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

Annotations already declare readOnlyHint=true, so the trailing "Read-only" is redundant, but the description earns credit for disclosing the score scale and its interpretation bands (100 identical, 90+ visually lossless, below 50 visibly degraded) and for noting both sizes are returned. It doesn't cover failure modes for invalid or mismatched inputs, so it stops short of a 5.

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?

It leads with the metric and its output, then the interpretation scale, then usage — a sensible front-loaded order with no wasted sentences. The scale brackets are dense but each band carries decision value; the redundant "Read-only" is the only filler.

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, the description correctly carries the burden of explaining the return value (SSIMULACRA 2 score plus both image sizes) and the schema covers the preview option. For a simple two-input read tool this is close to sufficient; only error/edge behavior is unaddressed.

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 description coverage is 100%, so the baseline is 3; the schema already documents original, candidate, the resize-on-size-mismatch behavior, and the preview thumbnail flag. The description adds no parameter detail beyond restating the return payload (both sizes).

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 names the exact operation (visual similarity comparison between an optimised image and its original) and the precise metric returned, SSIMULACRA 2, so an agent knows this is a quality-check tool rather than a generic analyzer. It doesn't explicitly contrast itself with the sibling analyze_images, leaving that differentiation to inference.

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?

"Use it to check results" implies the post-optimization verification workflow but never states when-not to use it or names an alternative sibling (analyze_images could plausibly be confused with it). The usage context is implied rather than spelled out.

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

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

Annotations already declare readOnlyHint=true and openWorldHint=false, so the safety profile is covered; the description adds genuinely non-obvious behavior: a bounded blocking wait, offset/nextOffset paging of results, and an in-memory retention window of one hour (or until server restart). It stops short of saying what happens when the job is not found or has expired, which is the one behavior an agent most needs to handle an error path.

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?

Three short sentences, front-loaded with purpose, then the wait behavior, then paging and retention. No filler, though the first sentence is a noun fragment that costs a beat of clarity compared to a verb-first phrasing.

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, the description carries the burden of telling the agent what comes back, and it does so at a high level (progress plus paged results, optional detailed mode implied by the schema). Retention and blocking semantics are the right things to disclose; only the expired/not-found behavior is missing.

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 only 40%, and the description covers just two of the five parameters: waitSeconds (already documented in the schema) and offset via the paging note. jobId, limit, and the concise/detailed response_format trade-off are left entirely to the schema, so it only partially compensates for the coverage gap.

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 resource (an optimize_images job) and what is returned (progress and results), so an agent can tell it apart from optimize_images, which creates the job. It doesn't explicitly name the sibling used to cancel or to fetch analysis, but the verb+resource pairing is unambiguous.

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?

The mention of waiting up to waitSeconds implies polling/blocking usage, and the phrasing hints this is the follow-up to optimize_images, but the description never says when to call it versus cancel_job or when a polling loop is appropriate. Usage is inferable rather than stated.

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

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

Annotations already declare destructiveHint=false and idempotentHint=true, but the description adds substantive context beyond them: new files are written to "<input folder>/optimized/" with originals untouched, re-runs skip unchanged files, there is a 45s execution ceiling after which a jobId is returned, and the allowed-directories line flags that no directories are currently configured. These are exactly the operational constraints an agent needs before invoking a write tool.

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 description is dense and front-loaded, leading with the default policy before enumerating variants and then the write/timeout behavior. It is on the long side and partially mirrors schema text (policies, resize semantics), but each sentence carries operational information rather than filler.

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 15-parameter, nested-schema, no-output-schema tool, the description covers the essentials: output location, idempotency, timeout/job handoff, dry-run preview, and the fact that no directories are configured. Minor gaps remain – urlBase, response_format, pagination (limit/offset), and metadata handling are left entirely to the schema, which does document them adequately.

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; the description earns an extra point by explaining how the parameters interact rather than restating them – per-image decisions, resizing respecting each image's own aspect ratio with no upscaling, widths producing srcset sizes, crop producing exact shapes. The policy tradeoffs are also given in prose with browser-compatibility context.

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 resource ("Optimise images") plus a scope qualifier ("deciding per image"), then enumerates the policies it chooses among. An agent can immediately separate it from analyze_images (read-only analysis) and get_job (job polling), which the description names explicitly.

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 clearly states the default behavior (policy "auto"), when to pick alternatives ("avif" for smaller files with Safari 16.4+, "picture" for fallback markup, "keep" to preserve format), and when to use dryRun ("to preview decisions without encoding"). It also routes to get_job when the run exceeds 45s. It does not explicitly say when to prefer this over analyze_images, so it falls short of a full 5.

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