Content
81%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 body is a model of actionable workflow design — hard gates with numeric pass criteria, corrective loops, tool assignments, and a complete worked example. Its weaknesses are efficiency-related: enumerations that duplicate the reference files inline, an orphaned saturation-signals.md reference, and behavioral constraints restated in two places.
Suggestions
Replace the inline enumerations (7 hook types, 8 content arcs in Phase 2 Layer 3; 6 saturation signals in Layer 2) with one-line pointers to the corresponding sections of references/virality-frameworks.md and references/saturation-signals.md, keeping only the names in SKILL.md — this cuts ~40 duplicated lines.
Wire the orphaned references/saturation-signals.md into the workflow: cite it in Phase 2 Layer 2 and Phase 4 validation (its own header claims it is 'Used in Phase 2 (layer 2) and Phase 4'), and reconcile the body's inline 6 signals with the file's 6 indicators so the two lists cannot drift apart.
De-duplicate the behavioral constraints: the role block, the Common Mistakes table, and the phase gates all restate Phone Camera Rule, 'BRIEFS not final content', and 'minimum 2 SCAMPER ops' — state each once at its point of enforcement and trim the rest.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | No generic-concept padding (it never explains what a reel or what SCAMPER is), but ~40 lines duplicate reference-file content inline: the 7 hook types and 8 content arcs restate sections of 'references/virality-frameworks.md', the 6 saturation signals restate 'references/saturation-signals.md', and the 6 Steal-Like-an-Artist rules and 5 validation checks restate 'references/adaptation-engine.md'. Role-block behavioral constraints (Phone Camera Rule, 'BRIEFS not final content', '2+ SCAMPER ops') are then repeated verbatim in the Common Mistakes table. Mostly efficient, but clearly could be tightened by pointing to the references instead of restating them. | 3 / 5 |
Actionability | Guidance is fully concrete and executable for an instruction-only skill: every phase names its tools (Glob/Grep/Read, WebSearch/WebFetch), exact domain file paths to read, numeric gates ('>= 10/18', 'minimum 5 references from at least 2 different niches AND 2 different countries', 'minimum 2 SCAMPER ops'), a table format to present findings, and a complete end-to-end worked example with the exact hook text ('Llevé un registro de TODO lo que comí por 30 días…'), CTA, posting time, and hashtags. Specific examples cover the common case, matching the copy-paste-ready anchor. | 5 / 5 |
Workflow Clarity | The five-phase pipeline is clearly sequenced with mode detection up front, a declared 'HARD GATE' in Phase 1, explicit GATE criteria between all subsequent phases, and corrective feedback loops ('If any check fails, apply corrective SCAMPER operation or discard', 'If not met, search more'). Phase 5 terminates in a 10-point final checklist (reference section 4), and the Common Mistakes table maps each error to its fix. This matches the anchor: explicit validation steps, error-recovery loops, and checklists. | 5 / 5 |
Progressive Disclosure | Four of five bundle files are cited at the right moments with verified section pointers ('references/research-sources.md sections 4.1 and 4.4', 'references/content-brief-template.md section 4', the adaptation-engine's five checks), all one level deep and all confirmed to exist with the cited sections. However, 'references/saturation-signals.md' is orphaned — SKILL.md never references it and instead inlines its own competing 6-signal list — so navigation to that file is impossible from the skill, and the inlined hook/arc enumerations duplicate virality-frameworks.md rather than deferring to it. Good structure with minor organization gaps, not the fully clean split of the top anchor. | 4 / 5 |
Total | 17 / 20 Passed |