Content
83%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 executable and well-structured: concrete commands, a real referenced script, pagination logic, and explicit success criteria. It is held back only by a slightly over-explained capability-boundary preamble and the absence of an explicit validate/retry checkpoint in the main extraction flow.
Suggestions
Tighten the 'Capability Components' preamble — the copy-paste/operational-boundary rationale can be one sentence instead of a paragraph.
Add an explicit post-extract validation step (e.g. 'if count == 0 or error == true, inspect page for updated .tF2Cxc selectors and retry') to close the feedback loop in the main flow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is largely efficient — command templates, a compact output example, and tight sections — though the 'Capability Components' preamble over-explains the copy-paste philosophy Claude could infer, keeping it just short of a 5. | 4 / 5 |
Actionability | Fully executable: concrete navigation URLs, a `wait stable` command, an exact `eval "$(python scripts/extract-social-profiles.py)"` invocation, and a complete JSON output example with field-by-field annotations. | 5 / 5 |
Workflow Clarity | Clear numbered sequence (navigate → wait → extract) with pagination handling and explicit success criteria; the batch guidance includes a 'test before batch' checkpoint and error resumption, but the core single-task flow lacks an explicit post-extract validation/retry loop, so it sits just below 5. | 4 / 5 |
Progressive Disclosure | Well-organized sections with a single real bundle script (scripts/extract-social-profiles.py) referenced inline and correctly invoked; the body stays overview-length rather than inlining the JS, with only minor organization gaps (no explicit pointer to memory-file contents). | 4 / 5 |
Total | 17 / 20 Passed |