Content
82%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, expert-level methodology document: concrete endpoints and flaw conditions, a genuinely useful chain table and decision tree, and disciplined non-destructive PoC framing with no wasted tokens. The main gaps are a handful of unexplained how-to-test bullets and no error-recovery guidance for paths that fail.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean throughout: compact bullet lists, concrete endpoints ('POST /api/users/me/email', 'PATCH /api/users/<id>'), and no padding or explanation of concepts Claude already knows. The one framing line ('ATO is an outcome, not a vuln class') is a genuinely useful heuristic, not filler — every token earns its place. | 5 / 5 |
Actionability | For an instruction-only methodology skill, the guidance is highly concrete: specific endpoints and parameters, specific flaw conditions ('TOTP code accepted from past N-step (clock skew abuse)', 'Token guessable (timestamp + user_id...)'), a concrete chain table, and an operational decision tree. It falls short of 5 because a few bullets state findings ('verification token leaks', 'Account recovery bypasses MFA entirely') without any indication of how to test or confirm them. | 4 / 5 |
Workflow Clarity | The decision tree gives an explicit sequenced workflow (MFA state → branch to specific numbered paths), and the PoC framing section acts as a validation checkpoint including the non-destructive proof requirement ('Show no destructive action taken on victim's actual data'). It is not a 5 because there are no feedback/recovery loops (e.g., what to re-examine when a path fails) and the path sections themselves are a catalog rather than an ordered procedure. | 4 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), and the ~95-line body is well-sectioned with numbered paths, a chain table, and clearly labeled cross-references to sibling skills that are one level deep and well signaled. It is not a 5 because the full methodology is inlined in SKILL.md — the per-path details and chain table could arguably live in reference files — and the body exceeds the under-50-lines simple-skill case. | 4 / 5 |
Total | 17 / 20 Passed |