Content
78%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 well-structured orchestration skill: concrete commands, per-subcommand step sequences, a genuine error-recovery table, and bundle references that match the real file tree exactly with disciplined on-demand loading. The only improvements are trimming the duplicated load-rule statements and clarifying what 'apply each prompt' should produce.
Suggestions
Consolidate the load rules: the Reference Files section repeats what the orchestration steps and runtime-context note already state — merge them into one authoritative section.
Clarify what 'Apply each prompt to the URL or topic' should produce (e.g., a structured finding per prompt) so the analysis step is fully executable without inferring output format.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence (no SEO concept explanations), but load rules are stated three times (runtime note, orchestration steps, Reference Files section) and the attribution notice appears twice. These minor duplications keep it just below the every-token-earns-its-place score-5 anchor. | 4 / 5 |
Actionability | Guidance is largely executable: exact file paths, the exact sync command ('${CLAUDE_PLUGIN_ROOT}/scripts/claude-seo' run sync_flow.py'), verbatim cross-reference strings, and an explicit 'Always surface exactly 2-3 prompts' rule. 'Apply each prompt to the URL or topic' delegates specifics to the prompt files, leaving a minor gap in what the output should look like. | 4 / 5 |
Workflow Clarity | Every sub-command has a clearly numbered sequence, and the Error Handling table provides explicit feedback loops (auth error → 'gh auth login' then retry; network error → check 'gh api rate_limit'). It lacks proactive validation checkpoints on analysis output, so it fits the score-4 anchor rather than the score-5 anchor with explicit validation steps. | 4 / 5 |
Progressive Disclosure | Verified against the actual bundle: all referenced paths exist and prompt counts match exactly (find=5, leverage=1, optimize=21, win=3, local=11 = 41). References are one hop deep, each with an explicit load condition, load-on-demand is mandated, and the optimize stage uses two-phase disclosure (list filenames, then load only 2-3 selected). This matches the score-5 anchor. | 5 / 5 |
Total | 17 / 20 Passed |