awslabs/aidlc-workflows v2.11.1-preview.20261010.1

AI-DLC Workflow 2.11.1-preview.20261010.1

on GitHub

Warning: Full Suite failed for this source. A Full Suite failure report ends these notes.

Merged commits since preview source b8d9bdc:

• test(input): the live check reads only what Codex said, and hands Codex the typed line whole (#2301)
• fix(sessions): a switch typed in another window asks this chat once instead of moving it (#2298)
• fix(kiro): the call handed to the agent carries no label for it to read back to the person (#2300)
• fix(engine): the gate line names no engine path, and the skill names itself AI-DLC, not a role (#2299)
• fix(hooks): the session-start context carries no heading the agent can read back to the person (#2297)
• test(input): the corpus check hands each hook its line as a file, so a loaded runner cannot record an empty turn (#2295)
• fix(engine): a lone letter is an exact pick only where the options were shown with letters (#2296)
• fix(construction): no step asks the agent for a session id (#2294)
• test(input): the input corpus through real agents, judged on what the person was told and what happened (live check + baseline) (#2292)
• fix(engine): the agent's commands act on this chat's work on Claude Code and Kiro CLI too (#2290)
• test(answers): what people reply to every question AI-DLC asks is read for meaning, and a tool records only an exact pick (corpus + unit check) (#2289)
• test(input): what people type after /aidlc, in 714 forms, is read the same way every time (corpus + unit check) (#2280)
• fix(engine): a chat keeps the work it is on, and a helper acts for the chat that spawned it on every harness (#2288)
• refactor(flags): one table of the flags next takes, read by the parser (#2287)
• fix(workspace): on Kiro IDE a bare intent name is looked up in this chat's space, and two intents with one name get their own answer (#2283)
• fix(flags): a flag AI-DLC cannot read never becomes the work name or a question about words the person never typed (#2281)
• fix(engine): a unit start in a Codex chat starts the unit instead of refusing for another chat (#2279)
• fix(workspace): a word after intent or space that is no verb and no record goes to the agent to read, words after a name are never dropped, and a re-entry report routes only the typed choice (#2276)
• fix(kiro-ide): the stage rules reach the agent again after a step reply was piped or captured away (#2167) (#2277)
• fix(codex): a review thread's engine command stays on the chat that started it (#2275)
• fix(review): the gate says when the reviewer was told its verdict, and the developer gets the whole brief (#2271)
• fix(orchestrate): an approved plan is not shown again as a stage gate, and the checkpoint remedy names a command the install has (#2272)
• fix(swarm): Retry after a failed landing keeps the approved plan instead of asking for it again (#2269)
• fix(approvals): keep reply recovery on the pending question (#2258)
• fix(codex): a helper agent acts on the chat that spawned it, not on the other open chat (#2270)
• fix(rules): a brief gets the stage's rules the one way step 2 says, rules_held included (#2257)

Source commit: cac57da

Nightly test report

Full Suite failed for cac57dada99e in run 38087105892. The preview build does not wait for these tests.

Failed legs: live_linux (failure), live_macos (failure), live_windows (failure).

Failing tests

• t-journey-question-once.sdk (1 failed assertions) in live-claude-sdk-2-macOS
  • t-journey-question-once (sdk): a question still open at the end of a turn is asked once > the routing question left open by Chat about this: nothing is recorded or asked again before the person answers, and their answer is taken
• t-tui-t27-depth-override.serial (1 failed assertions) in live-claude-tui-2-Windows, live-claude-tui-3-Linux

Runner errors

• live-claude-tui-2-Windows: INCOMPLETE: required selected coverage was not fully exercised
• live-claude-tui-3-Linux: INCOMPLETE: required selected coverage was not fully exercised

Failed jobs

• Linux / claude-tui (1): Linux / claude-tui 3/3
• Linux / result (1): Linux / result
• macOS / claude-sdk (1): macOS / claude-sdk 2/2
• Windows / claude-tui (1): Windows / claude-tui 2/3

Full logs are in the run 38087105892 full-suite-* artifacts.