Content
77%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, well-sequenced workflow skill: every phase has executable commands, explicit checkpoints, state persistence, and error-recovery loops. Its weaknesses are token efficiency — the full talk-script and beamer templates and duplicated color/rule tables bloat the body — and progressive disclosure, since large reference-worthy content is inlined while external references are unverifiable from the bundle.
Suggestions
Move the full TALK_SCRIPT.md markdown template (including the eight-entry Anticipated Q&A scaffold) and the complete beamer template into references/ files (e.g., references/talk-script-template.md, references/beamer-template.tex), keeping only a short example plus a one-line pointer in SKILL.md — this addresses both conciseness and progressive_disclosure.
Deduplicate content that is repeated between sections: the presentation rules in the Phase 2 table reappear nearly verbatim in 'Key Rules', and the Venue Color Schemes table is declared 'Same as /paper-poster-html' yet fully restated — reference the canonical table once instead.
Tighten or relocate the opening reviewer-routing blockquote: it is dense meta-procedural text about reviewer slot resolution that most invocations do not need inline; summarize it in one line and point to ../shared-references/reviewer-backends.md, and verify that path exists from this skill's installed location.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — it uses tables (Talk Type → Slide Count, Venue Color Schemes, presentation rules) and templates instead of prose, and does not explain concepts Claude already knows. However, there is real padding that could be tightened: the ~100-line TALK_SCRIPT.md markdown template with eight Q&A entries, the full beamer preamble, and the venue color table duplicated from '/paper-poster-html' are all inlined; presentation rules are stated twice (Phase 2 table and 'Key Rules'); and the opening reviewer-routing blockquote adds dense meta-procedural text only tangentially related to slide generation. This fits 'Mostly efficient but includes some unnecessary explanation or could be tightened' rather than the 4 anchor, where excess would be only minor. | 3 / 5 |
Actionability | The guidance is fully executable throughout: exact shell commands ('which pdflatex && which latexmk', 'cd slides && latexmk -$ENGINE -interaction=nonstopmode main.tex', symlink commands, the complete resolvable style-ref helper script with exit codes), a complete beamer template ready to adapt, a compile-error loop (max 3 attempts with log parsing), and pdfinfo page-count verification. Phase 7 specifies the PPTX script's exact behavior (aspect ratios, colors, notes transfer) rather than vague direction. This matches 'Fully executable; copy-paste ready code or commands; specific examples cover the common cases'; it is not a 4 because no key step lacks concrete detail. | 5 / 5 |
Workflow Clarity | Nine phases (0–8) are clearly sequenced, each ending with an explicit state write to SLIDES_STATE.json, and resume logic is defined ('if SLIDES_STATE.json exists with "status": "in_progress" and within 24h → resume'). Validation and feedback loops are present: compile error handling with a bounded retry loop, 'pdfinfo slides/main.pdf | grep Pages' verification against the outline, a mandatory user checkpoint after the outline ('⛔ STOP HERE and wait for user response'), and reviewer-unavailability fallback handling. This matches the 5 anchor 'Clear sequence with explicit validation steps; feedback loops for error recovery'. | 5 / 5 |
Progressive Disclosure | There is structure (constants, tables, per-phase sections, a final output summary), but no bundle files exist in this skill directory — the cross-references ('../shared-references/reviewer-backends.md', 'shared-references/integration-contract.md §2', 'tools/extract_paper_style.py') point outside the bundle and cannot be verified from it, i.e., references are not clearly signaled within the skill itself. Substantial content that would naturally be a reference file (the full TALK_SCRIPT.md template, the complete beamer template, the venue color tables) is inlined in a ~630-line body. This fits the 3 anchor 'Some structure but could be better organized; references present but not clearly signaled; content that should be separate is inline' rather than 4, where most content would be appropriately placed in separate files. | 3 / 5 |
Total | 16 / 20 Passed |