Skip to main content

AgentXchain v2.107.0

v2.107.0 closes the last behavior-proof gap across the social-posting wrappers and removes unnecessary fixture drag from the X/Twitter retry path.

Released: 2026-04-16

What Changed​

  • Added cli/test/linkedin-posting-script.test.js, a behavior-level fixture suite that executes the real LinkedIn posting wrapper with fake li-browser and python binaries across eight contract paths.
  • Added .planning/LINKEDIN_POSTING_TRUTH_BOUNDARY_SPEC.md so the LinkedIn wrapper's verified-success, fail-closed, and retry-boundary behavior is frozen as an explicit repo contract.
  • Patched cli/test/x-posting-script.test.js to replace the wrapper's sleep 5 with sleep 0 inside the fixture harness, preserving the fallback branch while cutting test wall clock from double-digit seconds to roughly 1.6 seconds.
  • Both social wrappers now have executable fixture proof instead of relying on grep-level content checks for the behavior that actually decides release-post truth.

Why This Matters​

Marketing automation is part of the shipped product surface. If a wrapper claims success without proof, the repo is inventing publication evidence. If the proof takes long enough that people stop running it, the repo has a different but equally real trust failure.

v2.107.0 fixes both sides of that problem. LinkedIn now has the same behavior-level wrapper proof standard as X/Twitter, and the X fixture no longer wastes real time on an artificial sleep that is irrelevant to the product contract.

That is the correct boundary: preserve the real control flow, remove the fake latency, and keep the proof cheap enough to run on every meaningful change.

Proof​

  • .planning/LINKEDIN_POSTING_TRUTH_BOUNDARY_SPEC.md
  • cli/test/linkedin-posting-script.test.js
  • cli/test/x-posting-script.test.js
  • cli/test/x-posting-truth-boundary.test.js
  • cli/test/marketing-browser-automation-content.test.js

Evidence​

  • 5069 tests / 1054 suites / 0 failures