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-engineered advisory skill body: highly actionable commands with explicit safety gates (confirmation before switchover, purchase-command ban, never fabricate output), clear routing to a well-organized one-level-deep reference bundle, and an Aurora/RDS confusion table that adds genuine non-obvious value. The main cost is token weight — the per-query "Critical Facts" scripts and restated constraints duplicate the reference workflows and make the body roughly 40% longer than it needs to be.
Suggestions
Move the "Critical Facts to Always Surface" per-query fact lists (upgrade advisor, MariaDB upgrade, RI vs DSP, Blue/Green VARCHAR→INT, PgBouncer vs Proxy) into their corresponding workflow references, keeping only a short body-level rule such as "always enumerate CLI command names, never mention LTS for RDS, never show purchase commands".
Delete or compress the instance-creation "Constraints" block that restates the 13 numbered best practices verbatim, and remove the hardcoded example model ID in favor of just the placeholder.
Add an explicit post-creation validation step (e.g. `aws rds describe-db-instances --db-instance-identifier <name>` and wait for `available`) to close the workflow-clarity checkpoint gap.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — no basic-concept explanations, concrete CLI commands, dense constraint lists — but it could be tightened considerably. The ~90-line "Critical Facts to Always Surface" section scripts mandatory fact lists for five specific example queries (e.g. "For '2x db.r7g.2xlarge RDS MySQL 24/7 — buy RI or Savings Plan?', you MUST tell the user ALL of the following seven facts"), duplicating content that belongs in the corresponding workflow references, and the instance-creation "Constraints" block restates the 13 numbered best practices immediately above it. Version-specific guidance ("prefer 8.4.x over 8.0", a hardcoded example model ID) is hedged with a check command but still adds time-sensitive detail. This fits the anchor 'mostly efficient but includes some unnecessary explanation or could be tightened' rather than the minor-trim level 4. | 3 / 5 |
Actionability | Guidance is fully executable throughout: exact CLI invocations with flags ("aws rds describe-db-engine-versions --engine <engine> --query ..."), a complete copy-paste create-db-instance example with all best-practice flags, post-creation TLS and KMS commands, and a named offline script with exact arguments ("--instance-type db.r7g.2xlarge --engine mysql --num-instances 2"). Specific examples cover the common cases, matching the top anchor; score 4 would require minor gaps that are not present. | 5 / 5 |
Workflow Clarity | The Common Tasks sequence (verify dependencies → classify and route via the Decision Guide → execute the workflow's Tasks section) is clear, and destructive operations have strong safeguards: prerequisite validation for Blue/Green ("binlog_format=ROW, automated backups enabled, instance in available state"), a mandatory user-confirmation gate before presenting switchover, post-switchover verification, and a hard ban on purchase commands. It falls short of the anchor-5 example because per-step validation detail is delegated to the reference files and the instance-creation workflow has no post-create verification step (e.g. describe-db-instances to confirm the instance reached 'available'), so some checkpoints are implicit rather than explicit in the body. | 4 / 5 |
Progressive Disclosure | Good structure against the actual bundle: a Decision Guide table routes each request class to one of four existing workflow references (all cited paths resolve, references are one level deep and never nest further), verify-dependencies and per-engine precheck files are well split, and a script is provided for offline pricing. The main gap is that the long "Critical Facts to Always Surface" section inlines per-example scripted answers that clearly belong in the corresponding workflow references, keeping this at 'most content appropriately placed, minor organization gaps' rather than the well-split top anchor. | 4 / 5 |
Total | 16 / 20 Passed |