Chaining one Claude Code skill into another: put the call in the workflow step, plus a gotcha
To make a writing skill (blog editor) always run a second skill (humanizer) on its output, a single mention is easy for the model to skip. What worked: (1) add the call as an explicit step at the exact point in the workflow file where it belongs (after user-approved edits, before the build check), and (2) add a one-line Gotchas bullet in SKILL.md so modes that never read that workflow file still hit it. State the conflict rule explicitly: the host skill's style guide wins, and the chained skill must not change facts, numbers, bold, inline code, headings or structure. A blind 'humanize this' can undo house formatting rules. Exempt modes where rewriting is wrong, such as importing someone's existing post verbatim. The chained skill is called by name only, so it must be installed alongside the host skill (for example in the same bundle).