The output of the /split-pdf skill looks very different here than the sample on your github as well as what the skill describes.
Is this purely due to the different prompt asking for the individual .md for each split etc, or do you have a modified version of the skill, too, that isn't on your github yet?
You know. I might have done that split-pdf before I did an update, but I thought it was an old skill that I hadn't updated in a while. So that's weird.
Okay, here is what Claude said when I relayed your question.
__
No — you haven't modified or pushed anything for `/split-pdf`. The repo is clean on that skill, and my deployed copy (`~/.claude/skills/split-pdf/SKILL.md`) is byte-identical to the one on GitHub. The only uncommitted items in mixtapetools are `.claude/worktrees/` and `claude/CLAUDE.html`, neither of which touch `/split-pdf`.
**So what your friend is seeing is explained by the prompt, not the skill.**
The `/split-pdf` SKILL.md on GitHub produces **one rolling file**: `articles/split_<name>/notes.md`, updated incrementally after each batch across eight structured extraction dimensions (research question, audience, method, data, statistical methods, findings, contributions, replication feasibility). That's the canonical output shape described on the repo.
What **we** produced during the CBS and Lu-Yu reads was different:
1. **One markdown per split** — `summaries/split-01_pp1-4.md`, `split-02_pp5-8.md`, etc., instead of one rolling `notes.md`.
2. **An `overall.md` / `overall_aej.md` synthesis file** at the end.
3. **Much longer per-split files** — dense equation derivations, not just structured extraction.
4. **A different folder layout** — `claudecode40/summaries/` and `claudecode40/summaries/lu_yu/`, not `articles/split_<name>/`.
All of that came from how you asked in the moment, not from a modified skill. Specifically, you told me:
- "Write a summary of each split in markdown. And then write a summary of the whole thing." (CBS, at the start)
- "Make the goal to deeply understand their TWFE estimation strategy, and a deep understanding of dose and outcomes in addition to whatever you ordinarily do." (Lu-Yu)
Both of those asks override the default `notes.md` pattern, so I wrote per-split files and an overall synthesis. The skill's core behavior — split into 4-page chunks, read 3 at a time, pause between batches — was followed; it's only the *output shape* that diverged, and that was prompt-driven.
**If you want `/split-pdf` to produce per-split summaries + an overall synthesis by default**, that's a reasonable enhancement to the skill and I can make it:
- Add a Q at the top: "single rolling `notes.md` (default) or one file per split + overall synthesis?"
You know, I didn't even think about it myself before either until you asked. I use the skills sort of like things that I get to then use sort of like however I want to. I just used beautiful_deck to review a deck, rather than make a deck, too. I'm not sure I do what I do correctly, but it does fit my style.
The output of the /split-pdf skill looks very different here than the sample on your github as well as what the skill describes.
Is this purely due to the different prompt asking for the individual .md for each split etc, or do you have a modified version of the skill, too, that isn't on your github yet?
Oh, it could be it didn't update. As I was writing, I was updating it, but maybe it didn't push. Let me check.
You know. I might have done that split-pdf before I did an update, but I thought it was an old skill that I hadn't updated in a while. So that's weird.
Okay, here is what Claude said when I relayed your question.
__
No — you haven't modified or pushed anything for `/split-pdf`. The repo is clean on that skill, and my deployed copy (`~/.claude/skills/split-pdf/SKILL.md`) is byte-identical to the one on GitHub. The only uncommitted items in mixtapetools are `.claude/worktrees/` and `claude/CLAUDE.html`, neither of which touch `/split-pdf`.
**So what your friend is seeing is explained by the prompt, not the skill.**
The `/split-pdf` SKILL.md on GitHub produces **one rolling file**: `articles/split_<name>/notes.md`, updated incrementally after each batch across eight structured extraction dimensions (research question, audience, method, data, statistical methods, findings, contributions, replication feasibility). That's the canonical output shape described on the repo.
What **we** produced during the CBS and Lu-Yu reads was different:
1. **One markdown per split** — `summaries/split-01_pp1-4.md`, `split-02_pp5-8.md`, etc., instead of one rolling `notes.md`.
2. **An `overall.md` / `overall_aej.md` synthesis file** at the end.
3. **Much longer per-split files** — dense equation derivations, not just structured extraction.
4. **A different folder layout** — `claudecode40/summaries/` and `claudecode40/summaries/lu_yu/`, not `articles/split_<name>/`.
All of that came from how you asked in the moment, not from a modified skill. Specifically, you told me:
- "Write a summary of each split in markdown. And then write a summary of the whole thing." (CBS, at the start)
- "Make the goal to deeply understand their TWFE estimation strategy, and a deep understanding of dose and outcomes in addition to whatever you ordinarily do." (Lu-Yu)
Both of those asks override the default `notes.md` pattern, so I wrote per-split files and an overall synthesis. The skill's core behavior — split into 4-page chunks, read 3 at a time, pause between batches — was followed; it's only the *output shape* that diverged, and that was prompt-driven.
**If you want `/split-pdf` to produce per-split summaries + an overall synthesis by default**, that's a reasonable enhancement to the skill and I can make it:
- Add a Q at the top: "single rolling `notes.md` (default) or one file per split + overall synthesis?"
- Document the alternative output shape
- Commit and push
thanks a lot for the quick and detailed reply!
I didn't realise a "manual override" of the standard parameters of the skill in the prompt can be so powerful and useful
You know, I didn't even think about it myself before either until you asked. I use the skills sort of like things that I get to then use sort of like however I want to. I just used beautiful_deck to review a deck, rather than make a deck, too. I'm not sure I do what I do correctly, but it does fit my style.