awslabs/aidlc-workflows v2.10.1-preview.20261005.1

AI-DLC Workflow 2.10.1-preview.20261005.1

on GitHub

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

Merged commits since preview source 53d09b1:

• fix(codex): the first question offers the engine's choices, and AI-DLC's files are never quoted to the person (#1817)
• fix(claude): only AI-DLC's own commands run with no prompt; a config or install change asks first (#1833)
• fix: keep update checks inside their budget by skipping attestation on version metadata (#1215)
• fix(config): a Claude Code refresh puts back AI-DLC's own settings entries and keeps the project's (#1300)
• fix(status): when the code was scanned, other open work, and only the checks someone switched off, in plain words (#1825)
• test(plan-approval): pin that a refactor or bugfix plan approved once gets built, through rule parts and a park (#1658)
• test(live drives): a drive fails when the agent records a decision the person did not make (#1798)
• feat(scope): per-scope collaborators switch to run stages lead-only (#1802)
• fix(stop-hook): a turn that ends at status, a setting or a scope change is not pushed on (#1830)
• fix(copilot): a refused AI-DLC command names what was not accepted and the form to run (#1730)
• fix(conductor): the task sidebar is used only when the session has its tools (#1835)
• fix(cursor): Cursor asks before running anything but AI-DLC's own workflow commands (#1823)
• fix(plugin): AI-DLC writes its own files only into real folders (#1841)
• fix(kiro): doctor and help keep their lines in the Kiro chat (#1839)
• fix(agents): AI-DLC's agents do file work with their file tools (#1787)
• fix(skill): only the person parks a workflow (#1838)
• fix(routing): a clone with work in progress names it, never "no workflow" (#1818)
• fix(config): a question with no answer, a wrong --from folder, and a runtime the terminal finds each say what to do (#1836)
• fix(stages): a named stage change says one line, not recompose's counts (#1837)
• test(live): t196 checks the person's named skips, and a drive ends when its finished subagents report inside the turn (#1840)
• fix: the person drives: the agent reads their reply, and the engine keeps their words (#1624)
• fix(dispatcher): input piped to a command reaches the tool in a copy (#1834)
• fix(kiro): the presence floor waits only on gates the person must answer (#1822)
• fix(learnings): a Unit's checkpoint offers one learnings question for the stages it walked (#1801)
• fix(kiro-ide): AI-DLC's engine commands run without an approval card on the copy channel (#1807)
• fix(engine): the gate names the real next stage and where the output is, and a skip says why (#1829)
• fix(stop-hook): a chat on work made from another chat ends without a push into its first stage (#1831)
• feat(config): choose the Kiro CLI session model in setup, and record no preset where a preset changes nothing (#1704)
• fix(routing): a setting typed with new work reaches the existing record the person picks (#1805)
• fix(replies): name the command the person types on their harness (#1784)

Source commit: 3f40ce6

Nightly test report

Full Suite failed for 3f40ce69c09d in run 37265765898. The preview build does not wait for these tests.

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

Failing tests

• t-guard-live-chat-lowering.sdk (1 failed assertions) in live-claude-sdk-1-macOS, live-claude-sdk-2-Linux, live-claude-sdk-2-Windows
  • production guards: lowering from a live Claude chat > typed switches lower the active intent's guards; plain words do not
• t-tui-t73-intent-capture.serial (1 failed assertions) in live-claude-tui-1-Linux
  • t-tui-t73-intent-capture (answering the stage gate produces artifacts on disk) > intent-capture jump commits intent-statement + answered questions on disk
• t122-stop-hook-e2e (1 failed assertions) in live-claude-sdk-1-Linux, live-claude-sdk-1-macOS, live-claude-sdk-2-Windows
  • t122 Stop hook end-to-end — real hook, real engine (sdk+cli) > (e2e) /aidlc --status over a completed workflow runs to done under the live Stop hook; done->allow trace asserted when the hook fires
• t183-codekb-placement-reverify.sdk (1 failed assertions) in live-claude-sdk-2-Windows
  • t183 codekb placement re-verify (sdk) — RE artifacts land at the engine-resolved space-level codekb > reverse-engineering writes its 9 artifacts to aidlc/spaces/<space>/codekb/<repo>/, NONE in the record dir

Failed jobs

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

Full logs are in the run 37265765898 full-suite-* artifacts.