[bitswell]

Summary

• New reference page plugins/loom/skills/loom/references/stacked-prs.md describing how the orchestrator publishes a read-only stacked-PR projection of an integrated LOOM epic via the stack-project tool.
• One-line addition to SKILL.md references list so the skill loader picks up the new page.

Companion to bitswell/loom-tools#25 which adds the stack-project tool itself. The recipe describes when and how the orchestrator should call the tool. This PR is doc-only.

Picked from the gh-stack RFP round — Team 1's "read-only presentation projection" angle.

Why this shape

The recipe deliberately emphasizes the load-bearing negatives in §5:

• Does not merge the stack (reviewer approvals are signals, not merges)
• Does not modify loom/<agent>-<slug> branches
• Does not touch workspace merge history
• Does not create new trailers, lifecycle states, or worker obligations
• Does not give reviewers a merge button

LOOM's --no-ff audit trail stays exactly as it is today. gh-stack is used as an output artifact, never as the integration mechanism. The "what this does NOT do" section is the single most important paragraph in the recipe — it's the reason the tool can exist without protocol changes.

Test plan

• Markdown renders cleanly
• Internal links resolve (protocol.md, examples.md)
• No references to the tool use shapes that diverge from the loom-tools PR
• Reviewer briefing — this recipe should be circulated to anyone who reviews stacked-mirror PRs, so they understand "approving the mirror is advisory, not a merge event"

🤖 Generated with Claude Code