Content
71%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.
The content is a well-structured, actionable workflow with strong sequencing, explicit validation, and a repair feedback loop, plus a concrete executable Chrome command and named verification tools. Its main weakness is conciseness, with repeated verbose restatements of the omit-internal-metadata rule and long conditional source-resolution clauses that could be tightened.
Suggestions
Consolidate the 'omit internal artifact/conversion metadata' guidance, which is restated nearly verbatim in step 1, step 4's negative check, the Standards section, and Verification Guidance; state it once authoritatively and reference it.
Tighten the long conditional clauses in step 1 (blob stub, sign-in page, redirect page, incomplete viewer, app shell) into a compact list of stop-conditions to reduce token cost without losing the gating logic.
If a concrete fallback renderer command is reliably available in the target environment, add an example command for the fallback path to close the minor actionability gap; otherwise the current tool-agnostic framing is acceptable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and task-focused (workflow steps, a concrete Chrome CLI block, standards bullets) but contains noticeable padding: long conditional clauses about blob stubs/sign-in pages/redirects and repeated, verbose restatements of the 'omit internal metadata' rule across the workflow and Standards sections. This fits 'Mostly efficient but includes some unnecessary explanation or could be tightened.' It is not a 4 because several sentences re-explain the same omit-internal-metadata guidance verbatim, and not a 2 because the core instruction is concrete and not padded with basic concept explanations Claude already knows. | 3 / 5 |
Actionability | Provides an executable, copy-paste-ready Chrome headless print-to-PDF command block and names concrete verification tools (pdfinfo, pdftotext, pdftoppm) with what each checks, matching 'Mostly executable guidance; concrete code or commands with minor gaps.' It is not a 5 because the fallback renderer path is intentionally tool-agnostic (no concrete command) and the verification examples are listed as 'or equivalent' rather than fully specified copy-paste commands, leaving minor gaps. | 4 / 5 |
Workflow Clarity | A clear six-step sequence (Resolve source -> Generate PDF -> Fallback -> Verify -> Repair -> Hand off) with explicit validation checkpoints: step 4 verifies the PDF, step 5 is a repair-before-handoff feedback loop ('fix... and regenerate'), and stop-with-blocker conditions are stated for source and renderer failures. This matches 'Clear sequence with explicit validation steps; feedback loops for error recovery.' It is not a 4 because validation is not merely present but looped (verify -> repair -> regenerate), and the destructive-output risk (handing off an unverified PDF) is explicitly gated. | 5 / 5 |
Progressive Disclosure | The body is a single self-contained SKILL.md with no bundle files (no references/, scripts/, or assets/ directories exist) and is organized into clear sections (Workflow, Standards, Verification Guidance). This fits 'Good structure; most content is appropriately placed... minor organization gaps' for a no-bundle skill. It is not a 5 because some content (the detailed verification tool list and the fallback-acceptable criteria) is inlined where a one-level reference could trim the overview, and not a 3 because the sections are well-signaled and not a monolithic wall of text. | 4 / 5 |
Total | 16 / 20 Passed |