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 tight, operationally complete playbook: every phase from identification through cracking to promotion and defender remediation is backed by concrete, executable commands with correct technical details (hashcat mode 18200, UAC bit filter, event 4768 detection tells). The only notable gaps are the non-executable kg_add_node notation for the promote step and the absence of explicit validation feedback loops, both minor.
Suggestions
Make the promote step executable by showing how kg_add_node is actually invoked (tool call syntax or script), matching the concreteness of the preceding sections.
Add a brief verification step after the hashcat run (e.g., check /tmp/asrep.hashes for cracked entries or run hashcat --show) to close the workflow with an explicit checkpoint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is command-dense and lean with almost no explanation of concepts Claude already knows, but minor editorializing could be trimmed ("sysadmin set DONT_REQ_PREAUTH to 'fix' a ticket issue in 2014, never reverted", "more powerful than kerberoast in some engagements"). This fits anchor 4 — efficient with minor over-explanation — rather than anchor 5 where every token earns its place. | 4 / 5 |
Actionability | Nearly all guidance is copy-paste ready: the ldapsearch command includes the exact UAC bitwise filter (1.2.840.113556.1.4.803:=4194304), both GetNPUsers.py variants have full flags, and hashcat -m 18200 includes the rule file. The one gap is the promote step, where kg_add_node(...) is call-signature notation rather than an executable command, fitting anchor 4 (minor gaps) instead of 5. | 4 / 5 |
Workflow Clarity | The numbered sections 1-5 give a clear identify → request → crack → promote sequence with alternative paths (hosted vs standalone tooling, zero-auth vs credentialed) and OPSEC considerations that function as decision checkpoints. There is no explicit validate/verify feedback loop, but the operations are non-destructive, so anchor 4 applies rather than the destructive-operation cap or anchor 5's explicit validation loops. | 4 / 5 |
Progressive Disclosure | There are no bundle files (references/, scripts/, assets/ are absent) and the body is a self-contained ~94-line playbook with clean numbered sections and no nested references. The size exceeds the under-50-line simple-skill exception, and everything is inline with no external splitting, fitting anchor 4 (good structure, content appropriately placed) rather than anchor 5's well-signaled reference structure. | 4 / 5 |
Total | 16 / 20 Passed |