Skip to main content

AgentXchain v2.106.0

v2.106.0 hardens the social-posting truth boundary so ambiguous X/Twitter submit outcomes are only reported as success when the post is actually proven on the account timeline.

Released: 2026-04-16

What Changed​

  • marketing/post-twitter.sh now verifies ambiguous-submit outcomes on both the primary attempt and the opposite-profile fallback attempt before it can exit successfully.
  • The X/Twitter verification matcher is now pipefail-safe, so a real timeline match no longer gets turned into a false negative by grep -q exiting early inside a shell pipeline.
  • Snippet extraction now trims leading and trailing whitespace in both post-twitter.sh and post-linkedin.sh, closing exact end-of-post false negatives in the shared verification pattern.
  • New executable shell coverage (cli/test/x-posting-script.test.js) proves the real wrapper contract across five paths: primary ambiguous verified, primary ambiguous unverified, fallback clean success, fallback ambiguous verified, and fallback ambiguous unverified.

Why This Matters​

Marketing automation is part of the public release surface. If a wrapper says a post succeeded when the post is still unverified, the repo is inventing evidence. That is the same quality failure as stale docs or overstated product claims.

This release fixes the boundary instead of papering over it. The X wrapper now has the same standard as the governed product itself: ambiguous execution is not success. It is only success after proof.

The whitespace trim matters for the same reason. A verifier that quietly appends a trailing space to the search snippet can miss a real post sitting right in front of it. That is not an edge case. That is a broken proof surface.

Proof​

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

Evidence​

  • 5061 tests / 1053 suites / 0 failures