You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
basecamp/basecamp-sdk#883 adds a Subtasks service on bc3's flat canonical subtask routes (bc3#12659). This surfaces all eight of its operations as a new basecamp subtasks group:
basecamp subtasks list <todo-or-card-id|url> [--limit N | --all | --page N]
basecamp subtasks show <id|url>
basecamp subtasks create <todo-or-card-id|url> <title> [--due DATE] [--assignees LIST]
basecamp subtasks update <id|url> [title] [--title T] [--due DATE | --no-due] [--assignees LIST | --no-assignees]
basecamp subtasks complete <id|url>
basecamp subtasks uncomplete <id|url>
basecamp subtasks move <id|url> --position N # 1-based
basecamp subtasks delete <id|url>
The routes are scoped to the account, so no verb takes --in. A parent is a to-do or a card, given by id or pasted URL. A subtask URL is its parent's URL ending in #__recording_<id>, and the fragment names the subtask.
update is partial, the way the SDK and bc3 are. --no-due and --no-assignees send the explicit clears ("due_on": "", "assignee_ids": []).
complete, uncomplete, move and delete get a bare 204 back, so their payload reports the state that was requested.
Registration, catalog entry, .surface, API-COVERAGE row, MCP model sync, skill docs, unit tests, e2e error-handling tests and a level-1 smoke suite all follow the shape of #677 (bubble-up), with #682 and #689 for the SDK-bump-plus-new-group pieces. The skill's old section on to-do subtasks, which used raw basecamp api calls, is replaced with the new commands.
⚠️ SDK pinned to main: swap for a release before merge
#883 has not been released yet, so go.mod pins basecamp-sdk main at the immutable commit 8216dd62c329 (v0.20.1-0.20260928111832-8216dd62c329), using make bump-sdk REF=<sha>. sdk-provenance.json, the vendored MCP model (PROVENANCE.json) and the Nix vendorHash all agree with that pin. Before merge, once basecamp-sdk cuts the release that ships Subtasks, re-pin to it (make bump-sdk REF=go/vX.Y.Z, scripts/sync-mcp-model.sh, make update-nix-hash).
What the pin brings along
Moving from v0.19.0 to main also brings in SDK changes the CLI had to absorb:
#930 split the Automation tag into Checkins, Templates, Webhooks, Lineup, Dock, Recordings and Search. The MCP automation domain now claims all seven, so the tools it serves are unchanged (the catalog snapshot diff is only the subtask actions). Subtasks joins the todos domain.
#924 checks event-feed refusal bodies when they are decoded. A body that isn't the documented shape now comes back as a statusless malformed response, which the SDK used to half-type. Four internal/mcpserver/feed_test.go cases assumed the old behavior. They now assert request_failed with no http_status, and the filter-mismatch fixture uses digests in the documented 16-hex shape. feed.go itself is unchanged.
#883 makes CardSteps.Reposition refuse 0. bc3 has always counted positions from 1, even though its docs said "zero indexed", so cards step move --position is now documented and validated as 1-based.
Test plan
bin/ci green locally
New unit tests (internal/commands/subtasks_test.go) cover each verb's route, method and body against the recording transport
e2e/subtasks.bats covers argument and flag errors
e2e/smoke/smoke_subtasks.bats (level 1) runs the full lifecycle against a real to-do on the next smoke run
basecamp/basecamp-sdk#883 models subtasks as a first-class resource on
bc3's flat canonical routes. This surfaces all eight operations as
`basecamp subtasks list|show|create|update|complete|uncomplete|move|delete`,
account-scoped (no --in), taking a to-do or card as the parent.
The SDK is pinned to basecamp-sdk main at 8216dd62c329 because #883 is not
released yet; the pin must be swapped for the release that ships it before
merge.
The pin carries three SDK changes past v0.19.0 that the CLI has to absorb:
- #930 split the Automation tag into Checkins, Templates, Webhooks, Lineup,
Dock, Recordings and Search. The MCP automation domain claims them all,
so its served surface is unchanged; Subtasks joins the todos domain.
- #924 refuses malformed event-feed refusal bodies at the decode as a
statusless malformed response. The MCP feed tests now assert that, and
the filter-mismatch fixture uses digests of the documented shape.
- #883 itself makes CardStepsService.Reposition refuse 0, since bc3 always
counted from 1, so `cards step move --position` is now 1-based.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds account-scoped subtask management across the CLI, SDK-backed MCP catalog, documentation, and test suites.
Changes:
Adds eight basecamp subtasks operations with URL support and pagination.
Refreshes SDK/MCP provenance and updates card-step positioning to 1-based.
Adds unit, e2e, and smoke coverage.
[!TIP]
If you aren't ready for review, convert to a draft PR.
Click "Convert to draft" or run gh pr ready --undo.
Click "Ready for review" or run gh pr ready to reengage.
…gative --limit
Review findings on #791:
- `subtasks delete` is a permanent, non-trashable delete, so it now goes
through ensureDeleteConfirmable and tui.ConfirmDangerous with --force,
the same gate `chat delete` uses.
- A to-do or card URL without a #__recording_ fragment names the parent;
the per-subtask verbs no longer read its id as a subtask id. Subtask and
card-step URLs, which name the subtask in the path, are accepted.
- `subtasks list --limit -1` is a usage error, as on the other lists.
- The skill's endpoint count follows API-COVERAGE to 203.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Codex and Copilot found two URL shapes subtaskIDArg still misread. One is a
subtasks or card-steps collection URL, which carries the parent's id. The
other is a numeric fragment on a URL that is not a to-do or card, which
includes a bare #<id> and a fragment on a subtask URL. A fragment now counts
only as #__recording_<id> on a to-do or card URL. A path id counts only on a
single subtask or step URL that has no fragment.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This summary renders 1 subtasks on #… for a single result, and the new unit test currently codifies that grammatical error. Select subtask/subtasks from the count and update the expected summary accordingly.
This branch has not been deployed
No deployments
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
github-actions
Bot
added
commands
What
basecamp/basecamp-sdk#883 adds a
Subtasksservice on bc3's flat canonical subtask routes (bc3#12659). This surfaces all eight of its operations as a newbasecamp subtasksgroup:--in. A parent is a to-do or a card, given by id or pasted URL. A subtask URL is its parent's URL ending in#__recording_<id>, and the fragment names the subtask.updateis partial, the way the SDK and bc3 are.--no-dueand--no-assigneessend the explicit clears ("due_on": "","assignee_ids": []).complete,uncomplete,moveanddeleteget a bare 204 back, so their payload reports the state that was requested.Registration, catalog entry,
.surface, API-COVERAGE row, MCP model sync, skill docs, unit tests, e2e error-handling tests and a level-1 smoke suite all follow the shape of #677 (bubble-up), with #682 and #689 for the SDK-bump-plus-new-group pieces. The skill's old section on to-do subtasks, which used rawbasecamp apicalls, is replaced with the new commands.main: swap for a release before merge#883 has not been released yet, so
go.modpins basecamp-sdkmainat the immutable commit8216dd62c329(v0.20.1-0.20260928111832-8216dd62c329), usingmake bump-sdk REF=<sha>.sdk-provenance.json, the vendored MCP model (PROVENANCE.json) and the NixvendorHashall agree with that pin. Before merge, once basecamp-sdk cuts the release that ships Subtasks, re-pin to it (make bump-sdk REF=go/vX.Y.Z,scripts/sync-mcp-model.sh,make update-nix-hash).What the pin brings along
Moving from v0.19.0 to
mainalso brings in SDK changes the CLI had to absorb:Automationtag intoCheckins,Templates,Webhooks,Lineup,Dock,RecordingsandSearch. The MCPautomationdomain now claims all seven, so the tools it serves are unchanged (the catalog snapshot diff is only the subtask actions).Subtasksjoins thetodosdomain.internal/mcpserver/feed_test.gocases assumed the old behavior. They now assertrequest_failedwith nohttp_status, and the filter-mismatch fixture uses digests in the documented 16-hex shape.feed.goitself is unchanged.CardSteps.Repositionrefuse0. bc3 has always counted positions from 1, even though its docs said "zero indexed", socards step move --positionis now documented and validated as 1-based.Test plan
bin/cigreen locallyinternal/commands/subtasks_test.go) cover each verb's route, method and body against the recording transporte2e/subtasks.batscovers argument and flag errorse2e/smoke/smoke_subtasks.bats(level 1) runs the full lifecycle against a real to-do on the next smoke runCard: https://app.basecamp.com/2914079/buckets/48521764/card_tables/cards/10346213842
🤖 Generated with Claude Code