City Atlas 城市圖譜

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

Beyond the read-only/idempotent annotations, the description discloses substantial behavioral caveats: per-metric year coverage (population 97–114, crashes 111–114, income 111→112, turnout by election cycle), the fact that 97–106 covers only certain counties, that pre-establishment figures for new neighbourhoods are estimates and excluded from comparison, that boundary-adjusted neighbourhoods are skipped with reasons listed, and that the tool reports magnitudes without causal claims.

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 core operation is front-loaded and every clause carries a constraint (year ranges, exclusions, threshold units), so little is filler. It is dense and parenthesis-heavy, which slows reading, but the length is justified by the number of metrics and caveats.

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 six-parameter comparison tool with no output schema, the description covers metric scope, data coverage limits, threshold semantics, and exclusion rules well. It does not describe the shape of the result beyond 'top increases/decreases,' leaving result interpretation mildly underspecified.

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 real value: it explains the default threshold values per metric (population 500, crash 5/20, income 5, turnout 3), states the unit of the threshold (people/cases/10k NT$/percentage points), and describes the valid year ranges per topic that constrain from/to.

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: compare two years and rank neighbourhoods by largest increase/decrease across population, crashes, income, and turnout. It names the reference implementation ('跟網站 07「看變化」同一套') and lists the exact metrics compared. It does not, however, distinguish itself from the sibling compare_groups, which an agent could easily confuse with it.

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 context is implied by the metric list and the note that thresholds are user-supplied, so an agent can infer this is for two-period difference ranking. But there is no explicit when-to-use versus compare_groups, rank, or need_ranking, and no statement of exclusions relative to those siblings.

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 cover the safety profile (readOnly, idempotent, non-destructive, closed-world). The description adds genuine behavioral context beyond that: the output is descriptive with no causal claims (只描述,不做因果) and the stratification exists to avoid city-center confounding. That interpretive caveat is valuable for how the agent frames results.

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 compact sentences, front-loaded with the core comparison before the stratification rationale. The bilingual restatement is mildly redundant but aids clarity; no sentence is wasted.

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 should characterize the return shape (per-stratum differences vs a single effect) and it does not. It does establish the analytical scope and a no-causation caveat, and annotations cover safety, but the agent lacks detail on what comes back.

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 schema already documents metric_a, has_b, and per. The description loosely maps A to metric_a and B to has_b and explains the stratum dimensions, but does not clarify how the 'per' conversion interacts with those strata. Baseline 3 is appropriate.

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 a specific analytical verb (compare A between cells with vs without B) and specifies the stratification design (residents × road length, 25 strata). It clearly conveys what the tool computes, though it doesn't explicitly position itself against siblings like compare_change (temporal) or rank.

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 the analytical framing but there is no explicit when-to-use, when-not, or named alternative. An agent can infer this is for cross-sectional stratified group comparison, but must reason about when to pick it over compare_change, rank, or overlay_metrics.

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, idempotentHint=true, destructiveHint=false, openWorldHint=false, so the safety profile is covered. The description adds value by disclosing the shape of the returned content (provenance fields, coverage, caveats), but says nothing about failure behavior for an unknown metric or any lookup side effects.

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 dense sentences that front-load the returned content, with no filler. The parenthetical enumeration is compact, though the trailing English sentence largely repeats the Chinese rather than adding new information.

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 read tool with no output schema and closed-world annotations, the description does the heavy lifting of sketching the return payload (provenance, coverage, statistics, caveats), so an agent knows what to expect. Only the ambiguity of the vague caveat phrasing and the absence of any error case keep it short of fully complete.

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?

With one parameter and 100% schema description coverage, the schema already documents 'metric' with examples (crash_ped4, 「行人事故」). The description only reinforces that it operates on a single indicator, adding no syntax or format detail beyond the schema.

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 a specific resource (one indicator / 一條資料) and enumerates what it returns: provenance (agency, dataset, URL, date, file, reconciliation), township coverage, statistics, and caveats. This is clearly a read/inspect tool, distinguishable in spirit from siblings like compare_*, rank, filter_cells, though no sibling is named 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?

There is no explicit when-to-use or when-not-to-use statement, nor any mention of alternatives such as search_catalog for discovery. The intent (inspect the provenance of a single known metric) is implied by the description but never framed as guidance.

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, idempotentHint, non-destructive, and closed-world, so the safety profile is covered. The description adds real behavioral context beyond that: the hard cap of four conditions, the grouping units (里/鄉鎮), and the fact that results are counts plus aggregated population rather than raw rows.

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 core capability and returns in two sentences, with the example placed where it clarifies rather than pads. Slightly diluted by restating the same content in Chinese and English, but no genuinely wasted text.

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 return-value burden and does describe the shape (matching cell count, population, aggregated by 里/鄉鎮). For a four-parameter filtering tool this is nearly sufficient; only the alternative-tool routing 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 description coverage is 100%, so the schema already documents top, town, unit, and the nested condition fields (metric, op, value) with enum meanings. The description reinforces the 'max four conditions' constraint but adds no syntax or semantics the schema lacks — baseline 3.

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?

Names a specific action and resource: find 125m grid cells satisfying up to four conditions, and states what is returned (cell count, population) and how it is aggregated (by 里/鄉鎮). The worked example ('離公車站超過 500 公尺、又住超過 100 人') makes the scope unambiguous, though it does not explicitly name the sibling it differs from (e.g. rank).

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 example implies a use case (multi-criteria filtering), but there is no explicit statement of when to use this versus rank, need_ranking, or compare_groups, nor any when-not guidance. 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 cover readOnly, idempotent, non-destructive and closed-world. The description adds real behavior beyond them: address matching is only resolved to a 125-meter grid, it uses the service's own address register, and it explicitly sends nothing to external services and records nothing — meaningful privacy/accuracy context that annotations do not convey.

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 accepted input types and the arrow-delimited output, so the key facts come first. The redundant English restatement at the end adds nothing, and the privacy sentence is dense, but overall it is compact and earns most of its space.

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 describing returns and does so: metrics plus ranking, plus the village-name extra fields. The 125-meter resolution limit on addresses is also disclosed. What is missing is pagination/limit behavior and any note on failure for unmatched inputs.

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 both parameters are already documented, including the 'default 10 metrics' behavior for metrics. The description adds example value formats for place (a full street address, a village name, a district name), which is mildly helpful but is essentially the same information the schema already carries.

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 accepted inputs (address, village name, township name) and the output shape (numbers plus a Greater Hsinchu ranking, with village-level extras), which is more specific than the title '查某個地方'. It is clearly a point/place lookup distinct from ranking and map siblings, but it never names an alternative tool by name.

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: you supply an address, village or district and get numbers back. There is no explicit 'use this when…' guidance, no exclusions, and no pointer to siblings like rank or show_map, so the agent must infer the boundary itself.

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, idempotentHint=true, and destructiveHint=false, so the safety profile is covered. The description adds the weighting scheme and its 60/20/10/10 default, which is useful context, but says nothing about output shape, units, or performance/rate characteristics beyond that.

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 short sentences with the core metric front-loaded; the weight-default note follows. The bilingual duplication adds mild redundancy but does not obscure the point, and there is no wasted preamble.

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 read-only ranking tool with no output schema, the description conveys what is being ranked and how the score is composed, which is enough for correct invocation. It would be stronger if it explained the output ordering or the unit of the percentile weighting.

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 `top`, `town`, `unit`, and the nested `weights` fields are already documented in the schema. The description's mention of the default weights restates schema content rather than adding syntax or meaning, so the baseline of 3 applies.

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 a concrete computed metric — where people lose the most time on public transport versus driving — and spells out its components (delayed person-minutes, jobs, schools, civic facilities, percentile-weighted). That is a specific verb+resource. However, it never distinguishes itself from the sibling `rank` tool, leaving ambiguity about which ranking entry point to pick.

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?

There is no explicit when-to-use, when-not-to-use, or alternative routing. With siblings like `rank`, `rank_...`, and `compare_groups` present, the agent gets no guidance on when this need-based ranking is the right choice. Only the weight defaulting rule is 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 readOnly, idempotent, non-destructive, and closed-world, so safety needs no restating. The description adds genuinely new behavioral context not covered by structured fields: the output is a nine-cell bivariate grid with counts per cell plus an explicit 'high on both' list, which is valuable since there is no output schema.

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 core concept and the return shape are front-loaded in the first clause, followed by a concrete example. It is dense and mixes Chinese and English, but no sentence is wasted.

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 7-parameter tool with no output schema, the description supplies the missing output shape (nine cells with counts and a high-both list), which is the key gap. It does not explain how the three-way split is computed (tertiles/quantiles) or how 'top' interacts with the grid, leaving a small but real gap.

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 every parameter including per, unit, shrink, and top is already documented in the schema. The description only hints at 'A (可換算)' and the 里/鄉鎮 unit, adding little beyond what the schema provides; baseline 3 applies.

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 states a specific analytic operation: two metrics each split into tertiles to form a 3x3 grid, returning per-cell counts of 里/鄉鎮 plus a list of places high on both. That is far more specific than a restatement of the name '兩樣疊'. It does not explicitly contrast itself with siblings like compare_groups or rank, so it stops short of a 5.

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 conveyed only by a worked example ('事故(每千居民)高、人行道少的里'), which implies when the tool is useful but never states when to prefer it over compare_groups, rank, or compare_change. No exclusions or prerequisites are given.

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 readOnly, idempotent, non-destructive, so the safety profile is covered. The description adds meaningful behaviour beyond that: the small-area shrinkage algorithm, that only the top N are returned, and that certain ranges are excluded from the result set.

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?

Dense but front-loaded, with the core ranking behaviour stated first and edge cases after. It is longer than strictly necessary in the Chinese portion, but each clause carries information and there is a compact English summary at the end.

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 covers the return shape (only the top N rows) and the metric-level and coverage constraints an agent needs. A few details, such as tie-breaking or what fields each returned row contains, remain unstated.

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 description clarifies some semantics (e.g. per can be a denominator code, filters example, unit restrictions), but much of it restates what the schema already documents, so it does not clearly exceed the baseline.

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 (rank) and resource (neighbourhoods/districts by an indicator), plus the normalisation behaviour, and flags the constraint on neighbourhood-level metrics. An agent can distinguish this from filter_cells or overlay_metrics without opening the 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 concrete conditions: it can restrict to one township, add filters, and only returns top N. It also states exclusions (neighbourhood-level data cannot be normalised; ranges without an assigned 里 do not enter the ranking). No sibling tool is named as an alternative, so it stops short of a 5.

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 cover read-only, idempotent, non-destructive and closed-world, so the bar is lower. The description still adds real context the annotations do not: the public server omits 4 Street-View-derived indicators, so only 157 of 161 are reachable, and it enumerates what each hit returns (code, name, unit, year, source, estimated flag, district coverage).

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?

Purpose is front-loaded, but the body is bloated with overlapping counts (11/82/161/141/20/47/4/80/157) and then largely repeated in English. Much of the numeric inventory is trivia that does not help an agent form a query.

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 usefully compensates by listing the returned fields and the coverage caveat. It is complete enough to call correctly, though it omits the natural next step (e.g. describe_metric for details), which is a notable gap given the sibling set.

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% and both parameters carry their own descriptions (limit default 12/max 40, query AND-matching). The prose adds nothing about parameter syntax or format beyond the schema, so the baseline 3 applies.

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 (search/找) and resource (catalog/資料目錄) and is unusually explicit about scope: 82 datasets, 11 categories, 161 indicators split into grid vs neighbourhood, plus elections. It stops short of a 5 because it never distinguishes itself from siblings like describe_metric, which an agent must pick between.

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 only implied via example keywords ('事故 行人', '公車', '淹水') and the note that whitespace-separated terms are ANDed. There is no statement of when to use this versus describe_metric or the compare_*/rank tools, and no exclusions, so the agent must infer the routing.

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 establish readOnly, idempotent, non-destructive, non-open-world behavior, so the description only needs to add rendering semantics. It does: drag/zoom and hover-for-numbers interactivity, plus an important fallback — non-rendering chats receive a clickable URL to pass to the user.

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?

Content is front-loaded with the core action, then the fallback behavior. The bilingual CN/EN pairing duplicates the same information, which is defensible for mixed-language agents but is not maximally economical.

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 explaining the return path, and it does so: an in-chat interactive map or, failing that, a link to hand to the user. Combined with full schema coverage, an agent has enough to invoke and relay the result correctly.

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 all five parameters are already documented in the schema. The description only loosely gestures at the inputs ('一條資料(或「需要度」)', '換算'), adding little beyond the enum and per-description detail the schema already carries.

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 output artifact: 'Draw an indicator as an interactive choropleth map of neighbourhoods', plus the granularity options (里/鄉鎮). It notes the equivalence to ranking ('換算同「排名」'), but does not explicitly contrast with siblings like overlay_metrics or compare_groups, so an agent must infer the boundary.

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 the rendering context ('在對話裡直接顯示') and by the ranking-equivalence note, which hints this is the spatial visualization counterpart to rank. There is no explicit when-to-use, when-not-to-use, or named alternative among the sibling tools.

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