Content
68%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-organized, largely actionable skill body with concrete workflows for its two primary commands. Its weaknesses are the undefined /legal-risk-assessment and /vendor-check workflows, a missing validation step for batch NDA triage, and a few lines of repo metadata that add tokens without operational value.
Suggestions
Add a validation checkpoint for batch triage: after processing, confirm the number of ratings matches the number of NDAs submitted and surface any that failed to parse before flagging items for human review.
Define the /legal-risk-assessment and /vendor-check workflows in their own sections (severity taxonomy critical/high/medium/low, escalation criteria, and data sources for vendor agreement status) instead of only referencing them in example prompts.
Trim or relocate the Source Repository and integrations lines (GitHub link, npx install, Slack/Box/Egnyte/Jira/M365 note) — they are repo metadata with no operational instructions for executing the commands.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean — per-command workflows, example prompts, and guardrails all earn their tokens — but the Source Repository section and the unactionable "Connects to Slack, Box, Egnyte, Jira, and Microsoft 365 when configured" line are minor padding that could be trimmed. Not 5: those sections add tokens without operational value; not 3: the padding is limited, not pervasive. | 4 / 5 |
Actionability | Concrete instruction-level guidance throughout: numbered steps ("Classify each issue: blocker / negotiate / accept"), a defined GREEN/YELLOW/RED output taxonomy, playbook intake questions, and worked example prompts showing expected inputs and outputs. Not 5: /legal-risk-assessment and /vendor-check are invoked ("Pair with /legal-risk-assessment", "/vendor-check [vendor name]") but never given their own workflow sections, leaving execution details unspecified. | 4 / 5 |
Workflow Clarity | /review-contract and /triage-nda have clear numbered sequences, but the skill explicitly supports batch operations ("Batch triage 5 NDAs... Rating per NDA; flag only those needing human review") with no validation checkpoint confirming every item was processed and rated — the rubric's batch-operation cap holds workflow clarity at 3. Not 4: the missing batch verification is exactly the checkpoint gap the cap targets; not 2: the core sequences themselves are well-defined. | 3 / 5 |
Progressive Disclosure | A ~72-line body cleanly sectioned per command with no external files needed, so no nested-reference or monolith problems; the only gaps are inline repo-metadata content and the two commands lacking sections. Not 5: it slightly exceeds the simple-skill threshold and those organization gaps exist; not 3: structure is good and nothing that clearly belongs in a separate file is inlined. | 4 / 5 |
Total | 15 / 20 Passed |