Content
75%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, highly actionable skill body: complete command coverage with fallbacks, a clearly sequenced workflow, and disciplined token use. It falls just short of top marks due to a non-executable pseudo-loop in the wiki step, absent validation for failed/empty results, and duplicated resolution logic between Steps 2 and 6.
Suggestions
Replace the pseudo-loop in Step 6 ("for each arxiv_id the user asked this skill to read in depth:") with an executable shell loop or a concrete per-ID command template so the entire wiki-ingest block is copy-paste runnable.
Add a validation checkpoint after Step 3 for empty or failed responses (e.g., if search returns no results, suggest /arxiv or /research-lit fallbacks) to close the workflow's feedback-loop gap.
Factor the duplicated ARIS_REPO/adapter resolution logic in Steps 2 and 6 into a single shared snippet or the referenced integration-contract file to remove the repeated block and tighten token use.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is almost entirely commands and workflow guidance that assume Claude's competence, with no explanations of known concepts; minor trims remain (the ARIS_REPO resolution logic is duplicated verbatim in Step 2 and Step 6, and the positioning table's prose partially repeats its message). Not 5 because of that duplication; better than 3 because there is no unnecessary explanation. | 4 / 5 |
Actionability | Every operation (search, brief, head, section, trending, web, sc) has executable adapter and raw-CLI fallback commands plus seven concrete invocation examples. Not 5 because the Step 6 wiki block contains a non-executable pseudo-loop line ("for each arxiv_id the user asked this skill to read in depth:"), so it is not fully copy-paste ready. | 4 / 5 |
Workflow Clarity | A clear six-step sequence with an explicit adapter-resolution checkpoint and fallback cascade, plus a well-defined depth escalation progression. Minor validation gaps remain: no handling of empty or failed search results and no post-ingest verification for the wiki step; the batch-write cap does not apply since ingestion is delegated to a helper that handles dedup with a warn-and-skip path. Not 5 due to the missing feedback loops on result failures. | 4 / 5 |
Progressive Disclosure | Well-organized sections (Role, Constants, Setup, Workflow, Key Rules) with two clearly-signaled, one-level-deep references to shared-references files, and no bundle files to split content into. Not 5 because the duplicated adapter/wiki resolution logic and setup detail could be factored into a referenced file, leaving minor organization gaps. | 4 / 5 |
Total | 16 / 20 Passed |