e28: Sync Pipeline Refactor — Parse→IR→Render

[danielvm-git]

Epic e28: Sync Pipeline Refactor

WSJF: 2.0 · BCP: 13 · Stories: 4
Source: specs/DEEPEN-ARCHITECTURE-REVIEW.md §5.1

(Renumbered from old e33 → current e28. Body updated to reflect current capsule.)

Refactor sync-skills.sh from copy-paste-variant to parse→IR→render pipeline. Extract a shared library, refactor the loop body into render-target functions.

Stories

#	BCP	Title
e28s01	3	Create scripts/lib/skill-common.sh with shared functions (resolve_repo_root, resolve_skills_root, parse_frontmatter, iterate_skills)
e28s02	2	Refactor 13 scripts to source skill-common.sh instead of duplicating REPO_ROOT/SKILLS_ROOT boilerplate
e28s03	5	Refactor sync-skills.sh to use render-target functions (one function per target: cursor, gemini-skill, gemini-command, pi-skill, pi-prompt)
e28s04	3	Refactor regenerate-lockfile.sh, generate-skill-index.sh, build-skill-index.sh to source skill-common.sh and use shared parse_frontmatter

Why

The current pipeline grew organically — each IDE target was copy-pasted with variations. Adding a new target (e.g., OpenCode) requires editing the loop body. With render-target functions, adding a target = adding a function.

Definition of Done

• bash scripts/sync-skills.sh produces identical artifacts as before (verified by golden suite G-04)
• 13 scripts share boilerplate from skill-common.sh
• Adding a new render target = writing one function

Current Status

backlog — not yet started. Capsule: specs/epics/e28-sync-pipeline/

[danielvm-git]

Closed as done — current e28 Sync Pipeline (renumbered from old e33). Landed in 2006f4b (PR #54): scripts/lib/skill-common.sh, render-target functions in sync-skills.sh, shared boilerplate across index scripts. All e28 story verifies pass.