feat(social): get_social takes a limit, and reputationEdges honours it #277

Closed

coilyco-ops wants to merge 1 commit from coilyco-ops/social-takes-a-limit into main AGit

coilyco-ops commented

2026-08-13 22:16:20 +00:00

Member

closes #286 - the tool accepted no limit, so its ~210-row reputationEdges had no bound it could honour

One of five pull requests against #286. PR 278 is stacked on this one, so land this first or land them together. Slice of #267, which stays open.

closes #286 - the tool accepted no limit, so its ~210-row reputationEdges had no bound it could honour One of five pull requests against #286. PR 278 is stacked on this one, so land this first or land them together. Slice of #267, which stays open.

coilyco-ops added 1 commit

2026-08-13 22:16:20 +00:00

feat(social): get_social takes a limit, and reputationEdges honours it 862b4248b3

get_social accepted no limit at all, so reputationEdges had no bound it
could honour: roughly 210 rows on a live server, one per trading pair,
growing with the community.

The parameter now appears in the tool's own schema. A bound a caller
cannot pass is not a bound they can use, and a test asserts the property
is declared rather than only implemented.

totalReputationTransfers still counts every transfer, which the test
pins: rule 5 of #267 protects the summaries, and bounding a ranked detail
array does not narrow them.

Deliberately untouched, each for its own reason:

  playByDay, firstLoginsByDay   time series. get_species thins these to
                                evenly-spaced samples with endpoints
                                kept, and a head slice would redefine the
                                curve as the first N days.
  reputationColumnsSeen         the exporter's header row, recorded so a
                                mis-named column is fixable from the
                                warning. Bounding it defeats its purpose.
  topReputationGivers,          sorted and never sliced, so both hold one
  topReputationReceivers        row per citizen despite the name. Same
                                finding as topBuyers in PR 276. Reported
                                on the issue rather than bounded here,
                                because this claim did not name them.

Mutations: dropping the bound fails the bounding test, and renaming the
schema property fails the declaration test.

Slice of #267, which stays open. get_map, get_world and get_climate still
take no limit, get_currency's holder cap is shadowed behind a cached
fetch, and the generic invariant test still wants a should-bound call per
array.

Refs #267

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-13 22:16:20 +00:00

coilyco-ops referenced this pull request

2026-08-13 22:16:40 +00:00

Invariant: limit must bound every unbounded array, and truncation must always warn — sweep all 25 MCP tools #267

coilyco-ops referenced this pull request

2026-08-13 22:17:20 +00:00

Invariant: limit must bound every unbounded array, and truncation must always warn — sweep all 25 MCP tools #267

coilyco-ops referenced this pull request

2026-08-13 22:38:41 +00:00

Four tools return arrays that limit never reaches, and one truncates silently #286