fix(pdf): read PDFs directly (1M context) instead of mandatory chunking#127
Merged
Conversation
Audit of all skills + hooks for model-staleness found 3 items, all in PDF handling — workarounds calibrated to the old ~200k context window: - pdf-processing.md: drop "Claude DOES NOT read directly", the mandatory 5-page Ghostscript split, "don't hold all chunks in working memory", and the 2-3-chunks-per-session cap. Default is now a direct Read (pages param, up to 20/request; a 1M-token window holds a full paper); Ghostscript page-range splitting is kept as a fallback for genuinely oversized/corrupt PDFs. - review-paper/SKILL.md: read directly and page via `pages`; the ~200k "budget escape" is now a configurable spend cap (~500k), decoupled from the context window (each review round already re-audits in fresh context). Everything else (hooks, effort:high pins, forked-context verification, max-round caps) was audited and confirmed current/intentional.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
A model-staleness audit of all skills + hooks (6-agent review) found the only genuinely outdated logic was in PDF handling — capacity workarounds calibrated to the old ~200k context window. Now that 1M context is the default and the Read tool reads PDFs natively, these are obsolete.
Fixed
pdf-processing.md— direct Read is now the default (pagesparam, up to 20/request; 1M window holds a full paper). Dropped "Claude DOES NOT read directly", the mandatory 5-page Ghostscript split, "don't hold all chunks in working memory", and the 2-3-chunks-per-session cap. Ghostscript/pdftk page-range splitting is kept as a fallback for genuinely oversized/corrupt PDFs.review-paper/SKILL.md— read directly + page viapages; the~200k"budget escape" is now a configurable spend cap (~500k), decoupled from the context window (each round already re-audits in fresh context).Audited and confirmed current/intentional (unchanged): all 6 hooks, the
effort: highpins (correct default, not escalated to xhigh), and forked-context verification + max-round caps (quality designs, not capacity workarounds).Verification
All three pre-commit gates pass (surface-sync · skill-integrity · model-versions). No count changes.