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 dense, command-first reference that is largely copy-paste actionable and free of padding. Its weaknesses are the absence of validation checkpoints in the token-theft-to-pivot flow, a few undefined variables and one half-finished code block (AAD Connect), and no progressive-disclosure split as subtopics accumulate.
Suggestions
Add a lightweight verification step after token theft, e.g. check `jq -r .access_token` is non-empty and probe `az rest`/ARM with a harmless GET before pivoting, plus guidance to enumerate assigned MI roles before requesting each audience.
Define or annotate the placeholder variables ($SUB, $VAULT_NAME, $TENANT, $CLIENT_ID) inline so every snippet is fully copy-paste ready, and either complete or remove the orphaned AesManaged line in the AAD Connect block.
Consider moving the pivot-pattern table, workload identity, and AAD Connect sections into one-level-deep reference files to keep SKILL.md a lean overview as the skill grows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean: executable commands, two compact tables, and one-line setup prose with no explanation of concepts Claude already knows. It falls short of the level-5 "every token earns its place" anchor due to the dead `$key = (New-Object System.Security.Cryptography.AesManaged).Key` line that serves no purpose and the redundant double attribution "MicroBurst (NetSPI), MicroBurst-Cloud (Karl Fosaaen)", which is trimmable noise rather than the padded explanation of the level-3 anchor. | 4 / 5 |
Actionability | Concrete copy-paste-ready curl/PowerShell covers Linux VM, Windows VM, App Service, ARM enumeration, Key Vault secret access, and the federated STS token exchange. Minor gaps keep it below level 5: `$SUB`, `$VAULT_NAME`, `$TENANT`, and `$CLIENT_ID` are used but never defined, and the AAD Connect block replaces real steps with a bare `Get-AADIntSyncCredentials` call assuming an unexplained module. | 4 / 5 |
Workflow Clarity | A logical sequence is implied by section order (steal token → pick audience → use token → pivot), matching the level-3 anchor of "steps listed but validation gaps". There are no checkpoints at all — no step to verify the token before use, no check of which MI assignments exist before requesting audiences, no error handling for failed token requests — so the level-4 anchor (most checkpoints present) does not fit. | 3 / 5 |
Progressive Disclosure | A single-file skill with clear section headers, well-organized tables, and no nested or dangling references; the ~95 lines of commands are appropriately kept inline. It sits at level 4 rather than 5 because distinct subtopics (pivot patterns, workload identity, AAD Connect, OPSEC) are moderately sized and could be split into one-level-deep reference files, and no reference structure is offered despite the content having grown past a simple single-purpose skill. | 4 / 5 |
Total | 15 / 20 Passed |