Content
88%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.
An excellent single-file diagnostic skill: fully executable commands with platform/permission fallbacks, a clearly sequenced workflow with validation guards and fall-through loops, and dense non-obvious detail that respects the reader's competence. The only improvements are trimming the twice-stated redaction and network-hang notes, and optionally moving the advanced stack-dump material to a reference file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious, earned tokens (BSD grep \b behavior, rss reported in KB, yama ptrace_scope, ps -u scoping) and assumes Claude's competence throughout, with no explanations of known concepts. It falls short of lean-perfection only through minor duplication: the credential-redaction note and the S-state/low-CPU network-hang guidance each appear twice. Anchor 4, not 3 (nothing pads or over-explains) and not 5 (the duplicated notes could be trimmed). | 4 / 5 |
Actionability | Every step is copy-paste executable with real flags and fallbacks: the ps enumeration pipeline, the node -e sidecar filter, the pgrep/tr/sed child listing, lsof -nP and ss -tnp variants, and the sample//proc stack-dump commands, each handling platform and permission differences (missing jq, gtimeout, CAP_NET_ADMIN). Fully executable guidance covering the common cases is the anchor-5 match. | 5 / 5 |
Workflow Clarity | The workflow is clearly sequenced (preamble resolving RUNTIME_DIR, a guarded PID fast path vs. the general 4-step path, then the report) with explicit validation checkpoints and feedback loops: kill -0 liveness guard, command-pattern refusal guard, digit-only argument validation, double CPU sampling, sidecar-to-ps fall-through, and ss-to-lsof fallback. The destructive/batch cap does not apply since the skill is explicitly diagnostic-only ("Don't kill or signal any processes"). | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are all absent) and the ~115-line body is well-sectioned (What to look for, Argument validation, Investigation steps, Report, Notes) with content appropriately placed for a single-purpose diagnostic. The under-50-lines simple-skill exception does not apply, and the advanced stack-dump section could arguably live in a separate reference file — minor organization gaps, matching anchor 4 rather than 5. | 4 / 5 |
Total | 18 / 20 Passed |