New experience pack ready for inclusion in the catalog.
Repo: https://github.com/anthroos/exp-publishing-an-openexp-pack
Author: ivan-pasichnyk
Slug: publishing-an-openexp-pack
Outcome label: pack_published at day_+0
Duration: 0 days · 48 steps (sub-day arc)
License: MIT
Shape
A meta-pack documenting the moment-to-moment publishing-and-coordination phase of another OpenExp pack: scrub-clean draft confirmed → public single-pack repo created → discovery topics added → catalog issue opened in upstream engine repo → local working dir cleaned up → two-tier convention rationalized → INDEX.md backfilled for the previously-published pack for symmetry → public repo cloned to local install path → mirror copies of public files removed from working dir → two internal PRs opened via git worktree off origin/main (private archive + narrow scrub skill that was authored earlier in the same session) → author pivots and asks for a broad authoring skill spun off from the same run → AI drafts long-form authoring SKILL.md → author overwhelmed by multi-PR checklist → AI simplifies to "I do all merges, you just run firebase deploy" → author approves "do it all" → background general-purpose agent delegated for mechanical site PR work → fourth PR opened for broad skill → three PRs merged sequentially → fourth PR conflicts on dispatcher (the just-merged scrub PR added a row in the same insertion zone) → AI rebases via worktree, resolves conflict by keeping both lines, force-pushes → fourth PR merges → AI runs firebase deploy directly → curl verification confirms site renders "2 packs available" → author proposes meta-pack documenting this very arc → AI runs step 0 of broad authoring skill on the meta-pack itself → author requests self-review → AI surfaces 6 issues including invented sub-day clock times → author picks rescope to honest publishing-and-coordination phase only → AI rewrites raw_timeline.
Pack format
Format v3 (raw): meta.yaml (facts only) + trajectory.anonymized.yaml (48 ordered steps, all on relative_day: 0) + README.md + SKILL.md. Author identity acknowledged in role tokens (<author>, <own_ai_assistant>, <own_delegated_agent>); resource names abstracted to category tokens (<single_pack_repo>, <upstream_engine_repo>, <site_repo>, <site_project_id>, etc.). No applies_when, no summary, no author grade.
Why this might be the most useful pack of the three
The first two packs (Norda, FullSwing) are sales arcs — useful for sellers running similar acquisition motions. This one is about the publishing pipeline itself — useful for anyone who wants to publish their first OpenExp pack. It is the lead-magnet pack: install it, read the trajectory, see what shipping a pack actually looks like in real time, including friction points (multi-PR conflict, two-tier convention, on-the-fly skill distillation, delegation decisions).
Add to catalog
Please add to openexp.ai/use alongside the existing two packs by the same author.
New experience pack ready for inclusion in the catalog.
Repo: https://github.com/anthroos/exp-publishing-an-openexp-pack
Author: ivan-pasichnyk
Slug: publishing-an-openexp-pack
Outcome label:
pack_publishedatday_+0Duration: 0 days · 48 steps (sub-day arc)
License: MIT
Shape
A meta-pack documenting the moment-to-moment publishing-and-coordination phase of another OpenExp pack: scrub-clean draft confirmed → public single-pack repo created → discovery topics added → catalog issue opened in upstream engine repo → local working dir cleaned up → two-tier convention rationalized → INDEX.md backfilled for the previously-published pack for symmetry → public repo cloned to local install path → mirror copies of public files removed from working dir → two internal PRs opened via git worktree off origin/main (private archive + narrow scrub skill that was authored earlier in the same session) → author pivots and asks for a broad authoring skill spun off from the same run → AI drafts long-form authoring SKILL.md → author overwhelmed by multi-PR checklist → AI simplifies to "I do all merges, you just run firebase deploy" → author approves "do it all" → background general-purpose agent delegated for mechanical site PR work → fourth PR opened for broad skill → three PRs merged sequentially → fourth PR conflicts on dispatcher (the just-merged scrub PR added a row in the same insertion zone) → AI rebases via worktree, resolves conflict by keeping both lines, force-pushes → fourth PR merges → AI runs firebase deploy directly → curl verification confirms site renders "2 packs available" → author proposes meta-pack documenting this very arc → AI runs step 0 of broad authoring skill on the meta-pack itself → author requests self-review → AI surfaces 6 issues including invented sub-day clock times → author picks rescope to honest publishing-and-coordination phase only → AI rewrites raw_timeline.
Pack format
Format v3 (raw):
meta.yaml(facts only) +trajectory.anonymized.yaml(48 ordered steps, all onrelative_day: 0) +README.md+SKILL.md. Author identity acknowledged in role tokens (<author>,<own_ai_assistant>,<own_delegated_agent>); resource names abstracted to category tokens (<single_pack_repo>,<upstream_engine_repo>,<site_repo>,<site_project_id>, etc.). Noapplies_when, no summary, no author grade.Why this might be the most useful pack of the three
The first two packs (Norda, FullSwing) are sales arcs — useful for sellers running similar acquisition motions. This one is about the publishing pipeline itself — useful for anyone who wants to publish their first OpenExp pack. It is the lead-magnet pack: install it, read the trajectory, see what shipping a pack actually looks like in real time, including friction points (multi-PR conflict, two-tier convention, on-the-fly skill distillation, delegation decisions).
Add to catalog
Please add to
openexp.ai/usealongside the existing two packs by the same author.