Content
70%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with a well-sequenced, validated workflow and copy-paste-ready code for every stage. Its main weaknesses are some repeated guidance that could be tightened and a set of prompts/ references that point at files not present in the bundle, which breaks progressive disclosure.
Suggestions
Add the missing prompts/ directory (style_from_query.md, style_from_image.md, outline.md, page_prompt.md) or inline the prompt scaffolding so the referenced sys-prompt files actually exist.
De-duplicate the call-routing policy (it appears as a table, inline prose, and Hard Rules #3/#4) into a single authoritative section.
Consider moving the long negative-prompt string and Stage 5/6 behavioral notes into a references/ file to reduce the inline weight while keeping the SKILL.md overview navigable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely efficient and assumes competence (executable inline python3 -c blocks, terse tables), but it repeats routing rules across the routing table, the Hard Rules, and inline caveats, and includes some restated guidance ('Never mix…') that could be tightened, matching the level-2 'mostly efficient but could be tightened' anchor. | 2 / 3 |
Actionability | Provides fully executable, copy-paste-ready code and commands for every stage (python3 -c blocks importing model_client, sn_agent_runner.py invocations, build_pptx.py, resume_scan.py), each with concrete flags and example output, matching the level-3 'fully executable; copy-paste ready' anchor. | 3 / 3 |
Workflow Clarity | Stages 2–6 are explicitly sequenced with a resume-scan dispatch table, per-page action dispatch, explicit validation checkpoints (assert page_count, JSON parse guards, resume re-entry), and clear failure/feedback handling ('record, echo, continue'; re-validate via resume), matching the level-3 clear-sequence-with-validation anchor. | 3 / 3 |
Progressive Disclosure | The body references $SKILL_DIR/prompts/style_from_query.md, style_from_image.md, outline.md, and page_prompt.md as essential sys-prompt sources, but no prompts/ directory exists in the bundle — the references are dangling rather than one-level-deep and clearly signaled, matching the level-1 'poorly organized / missing referenced files' case. | 1 / 3 |
Total | 9 / 12 Passed |