Skip to content

fix: BUG-001: workflow success can hide missing task evidence and failed agent execution - #136

Open
andrei-hasna wants to merge 1 commit into
mainfrom
factory/1db49c9f-63d5-422c-959b-da16a65c-0aaa3977
Open

fix: BUG-001: workflow success can hide missing task evidence and failed agent execution#136
andrei-hasna wants to merge 1 commit into
mainfrom
factory/1db49c9f-63d5-422c-959b-da16a65c-0aaa3977

Conversation

@andrei-hasna

@andrei-hasna andrei-hasna commented Jul 27, 2026

Copy link
Copy Markdown
Contributor

Objective

BUG-001: workflow success can hide missing task evidence and failed agent execution

Observed while launching platform-socializer on 2026-06-30.

Affected Socializer route: socializer-launch-backend-ready-drain (loop d449c31542ba), template todos-task-worker-verifier, provider codewith, worktree-mode required.

Symptoms:

  • Workflow 786e16b5c0e9 for task 318fc023-58dd-47a0-b94e-6997cd9ce469 is marked succeeded, but adversarial audit reported worker/verifier shell execution was blocked before simple commands could run; Todos task remains pending and has no evidence comments.
  • Workflows 8290b6d405d1 / eda80e8a and 93f937c12ced / f8919326 reached worker+verifier succeeded but left weak or missing task-level evidence.
  • Route work-item state shows succeeded/admitted even when the task remains pending, which lets the ready queue continue while launch gates are not actually proven.

Expected:

  • Worker/verifier agent failures or blocked tool execution must fail the workflow or mark the route item failed, not succeeded.
  • The template should persist concise worker/verifier stdout summaries or evidence paths, and/or enforce a final task evidence gate before route success.
  • Completion should require independent verifier completion and visible task evidence.

Current workaround:

  • Paused Socializer backend drain d449c31542ba. UI lane remains active because c49bef80 has visible worker and independent verifier evidence.

Do not include secrets. Repro evidence is in /home/hasna/.hasna/loops/runs/platform-socializer and task comments for f8919326, 318fc023, 680d728f, 9ca7457e, f1b34411, eda80e8a.

Verification

  • policy source: base 35c9ec6 (immutable commit — agent-proof)
  • install: pass
  • typecheck: pass
  • build: pass
  • test: pass

Run run_93bb8a518732 · backend codewith · task 1db49c9f-63d5-422c-959b-da16a65c7718
🏭 Generated by @hasnaxyz/factory


View with [code]smith Autofix with [code]smith
Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant