Sync workflow regenerates CHANGELOG.md with a policy that differs from the caller's changelog:check #226

Problem

The reusable pm-github-sync.yml regenerates CHANGELOG.md after importing issues. When the caller keeps the default post-import-script (changelog:full), the workflow appends --no-pending-release. The caller's own changelog:check (run by CI on the sync PR) uses the package's normal release policy, so the regenerated file has a different heading and every gated sync PR fails after its runtime tests pass.

Observed on 2026-11-07/08 in pm-http#8, pm-hypothesis#14, pm-review#70, pm-rlm#57 and pm-rl#130.

Current workaround (per caller)

The callers now set post-import-script: "changelog:automation", a package script that delegates to the unchanged changelog:full. The workflow injects the flag only for its default script name, so this sidesteps it:

• Keep issue-sync changelog generation consistent with CI pm-http#22
• Keep issue-sync changelog generation consistent with CI pm-hypothesis#23
• Keep issue-sync changelog generation consistent with CI pm-review#76
• Keep issue-sync changelog generation consistent with CI pm-rlm#60
• Keep issue-sync changelog generation consistent with CI pm-rl#159

Every other caller with the default script still hits the same failure.

Expected

A sync PR's regenerated changelog must pass the caller's own changelog:check without per-caller configuration. Either:

• stop injecting --no-pending-release, and run the caller's script exactly as declared; or
• regenerate with the same flags the caller's changelog:check uses.

Add a contract test in which a fixture caller's sync output passes its changelog:check before and after a release tag exists. Once the fix ships, the per-caller changelog:automation inputs can be removed.