Content
20%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a verbose persona description padded with generic domain knowledge Claude already has, offering little concrete or executable guidance. It has a sequenced but checkpoint-less workflow and is a monolithic inline document rather than an overview pointing to reference materials.
Suggestions
Cut the Capabilities, Behavioral Traits, and Knowledge Base sections down to the non-obvious specifics Claude would not already know, and move vendor/tool detail into a reference file.
Replace abstract Instructions and Response Approach steps with concrete, executable procedures (e.g., specific ticket-routing rules, a CSAT analysis template, a checklist with explicit validation steps).
Add a validation/feedback checkpoint to the Response Approach (e.g., 'Confirm resolution with the customer before closing; re-measure CSAT after follow-up') and split the bulk content into referenced sub-files for proper progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~150-line body enumerates generic domain knowledge Claude already knows (vendor lists like Salesforce/HubSpot/Zendesk, definitions of CSAT/NPS, 'Empathy-first approach', 'Modern customer support platforms'), padding the context window with material that earns no new capability, matching the 'verbose; explains concepts Claude knows; padded with unnecessary context' anchor. | 1 / 3 |
Actionability | Guidance is abstract throughout ('Clarify goals, constraints, and required inputs', 'Identify the best solution using available tools') with no concrete commands, code, or specific procedures; the Capabilities section describes a persona rather than instructing actions, matching 'vague or abstract; no concrete code/commands; describes rather than instructs.' | 1 / 3 |
Workflow Clarity | The Response Approach provides a clear 10-step sequence (listen -> analyze -> identify -> communicate -> verify -> follow up -> document -> optimize -> escalate -> measure), so a sequence exists, but there are no validation checkpoints, decision criteria, or feedback loops, fitting 'steps listed but validation gaps; checkpoints missing or implicit.' | 2 / 3 |
Progressive Disclosure | The body uses section headers for some structure, but it is a monolithic inline wall of capability/knowledge lists with no external reference files (and none exist in references/scripts/assets), so content that should be split into reference files is inline, matching 'some structure but could be better organized; content that should be separate is inline.' | 2 / 3 |
Total | 6 / 12 Passed |