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
{{ message }}
Repository navigation
[Bug]: The Human criterion shown on the inception page cannot be ticked there, and nothing says where it can #93
The inception page (/inception/T-XXX) is where the operator is sent to decide. It shows the task's Acceptance Criteria, including the ### Human criterion, as plain text: <li>[ ] [REVIEW] ..., with no input. The operator sees a box, clicks it, and nothing happens.
The working checkbox exists on /tasks/T-XXX (and /approvals) only. The inception page does not link to /tasks/T-XXX; its only link to those places is the generic /approvals navigation entry.
Consequence observed on 1.8.3: the operator recorded GO from the inception page, which shows the objectives he was approving. The decision was saved, but the [REVIEW] Human criterion of the same task stayed unticked, so fw inception sweep answered "ticked but 1 Human AC still unchecked — stays in active/" and the task could not leave active/. Once the operator found /tasks/T-XXX and ticked the box there, the sweep completed the task.
Steps to Reproduce
Open /inception/T-XXX for an inception task that has an unticked ### Human criterion.
Try to tick the criterion in the "Acceptance Criteria" card.
Look for a link to the page where it can be ticked.
Expected Behavior
Either the criterion can be ticked on the page where the decision is taken, or the page (and the "decision is saved" message) says that one more click is needed and links to /tasks/T-XXX.
Actual Behavior
On the served inception page (1.8.6): no type="checkbox" input, no toggle-ac request target, no href="/tasks/T-XXX". The criterion is rendered as:
<li>[ ] [REVIEW] Review exploration findings and approve go/no-go decision
On /tasks/T-XXX for the same task, the checkbox inputs and the toggle-ac targets are present.
fw doctor output
AI Agent
Claude Code
Operating System
Linux
Additional Context
Version: 1.8.6 (1ea1044), vendored consumer. The rendering was checked on the served pages of 1.8.6. The consequence on the decide chain (task stuck in active/) was observed on 1.8.3 and not replayed on 1.8.6.
web/templates/inception_detail.html, lines 360 to 364: {{ sections.acceptance_criteria }} in a plain section card.
Suggestions:
Render the Human criteria of the inception page with the same working checkboxes as the task page.
Or link each criterion, and the "decision is saved" message, to /tasks/T-XXX.
Thank you. This issue has been ingested into the Agentic Engineering Framework's intake as INTAKE-0009. Next step: classification (type, affected area, severity, whether it reproduces on the current release and on bleeding-edge). You will get a comment here at each step: classified → decision → in progress → fixed on bleeding-edge → available for testing → released.
These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.
Fixed under T-4046 in commit 770394b (on the bleeding-edge branch; not yet in a release). Both your suggestions are in:
The inception page's "Acceptance Criteria" card now renders each criterion with a working checkbox that posts to the same toggle-ac endpoint the task page uses (same CSRF, same provenance record; there is no second way to tick). Human criteria keep their Steps / Expected / If not.
The card links to /tasks/T-XXX, and the "Decision recorded — one step left" card names any Human criterion still open and links to it, so a GO recorded on this page no longer leaves the task stuck in active/ without saying why.
Regression test tests/web/test_t4046_inception_human_ac_tickable.py: both tests fail on the old code; the decide end-to-end tests stay green. Running on our own Watchtower now; the maintainer's visual check is pending.
To test it before the release:git clone -b bleeding-edge https://github.com/DimitriGeelen/agentic-engineering-framework.git aef-be && aef-be/bin/fw upgrade /path/to/your/project, then fw watchtower restart and open /inception/T-XXX for an inception with an open Human criterion. Feedback here is welcome. Next update here: released.
These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.
Activity
DimitriGeelen commentedon Oct 10, 2026
[AEF intake] Received — 2026-10-10
Thank you. This issue has been ingested into the Agentic Engineering Framework's intake as INTAKE-0009. Next step: classification (type, affected area, severity, whether it reproduces on the current release and on bleeding-edge). You will get a comment here at each step: classified → decision → in progress → fixed on bleeding-edge → available for testing → released.
These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.
DimitriGeelen commentedon Oct 10, 2026
[AEF intake] Classified — 2026-10-10
[AEF intake] Accepted — 2026-10-10
Tracked as T-4046, prioritised (horizon: now). Next update here: in progress.
These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.
T-4046: scope GitHub #93 — inception page criteria tickable through t…
T-4046: inception page criteria tickable via shared _ac_list partial …
DimitriGeelen commentedon Oct 11, 2026
[AEF intake] Fixed on bleeding-edge — 2026-10-11
Fixed under T-4046 in commit 770394b (on the
bleeding-edgebranch; not yet in a release). Both your suggestions are in:toggle-acendpoint the task page uses (same CSRF, same provenance record; there is no second way to tick). Human criteria keep their Steps / Expected / If not./tasks/T-XXX, and the "Decision recorded — one step left" card names any Human criterion still open and links to it, so a GO recorded on this page no longer leaves the task stuck inactive/without saying why.Regression test
tests/web/test_t4046_inception_human_ac_tickable.py: both tests fail on the old code; the decide end-to-end tests stay green. Running on our own Watchtower now; the maintainer's visual check is pending.To test it before the release:
git clone -b bleeding-edge https://github.com/DimitriGeelen/agentic-engineering-framework.git aef-be && aef-be/bin/fw upgrade /path/to/your/project, thenfw watchtower restartand open/inception/T-XXXfor an inception with an open Human criterion. Feedback here is welcome. Next update here: released.These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.
DimitriGeelen commentedon Oct 18, 2026
[AEF intake] Released — 2026-10-18
T-4046 is released in v1.8.11.
Please confirm it fixes this for you. Reply here if it does not. With no reply in 14 days this issue is closed with a note, and you can reopen it.
How it was tested:
tests/web/test_t4046_inception_human_ac_tickable.py(2 tests: the page has a tickable Human criterion; the existing toggle endpoint ticks the box). Green on bleeding-edge.Why it is released now: v1.8.11 passed the full-suite comparison (https://github.com/DimitriGeelen/agentic-engineering-framework/blob/50fae848ff46e156c1be630e85b0644aa070be38/docs/reports/T-4246-full-suite-comparison.md) and the operator's release decision.
Field evidence: none. The page layout awaits the maintainer's review. Please confirm this works for you.
Release: https://github.com/DimitriGeelen/agentic-engineering-framework/releases/tag/v1.8.11 · Field report: https://github.com/DimitriGeelen/agentic-engineering-framework/blob/50fae848ff46e156c1be630e85b0644aa070be38/docs/reports/field-test-v1.8.11-results.md
These status updates are automated, but Dimitri personally looks at every issue and decides what happens to it. Replies on this issue are read.