Content
65%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The content is highly actionable with concrete commands across a well-sequenced five-phase workflow, but it is monolithic: it re-explains known concepts, carries generic boilerplate and an erroneous Python prerequisite, and entirely ignores the provided bundle files. Tightening the overview and routing detail to the existing references would lift the weaker dimensions.
Suggestions
Link the existing bundle files from the body — e.g., point detailed listener/implant syntax to references/api-reference.md, posture/OPSEC notes to references/standards.md, and per-phase runbooks to references/workflows.md — so SKILL.md becomes an overview rather than a monolith.
Trim the Overview's re-explanation of what Sliver is and replace the generic 'When to Use' boilerplate; remove the 'Python 3.8+' prerequisite, which is incorrect for a Go-based framework.
Insert inline validation checkpoints inside each phase (e.g., verify the team server is listening after 'systemctl start sliver', confirm the redirector proxies before generating implants) with explicit fix-and-retry guidance, rather than only a detached end-of-document checklist.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly command-driven, but the Overview re-explains what Sliver is ('an open-source, cross-platform adversary emulation framework ... written in Go'), the 'When to Use' section is generic auto-generated boilerplate, and Prerequisites oddly lists 'Python 3.8+' for a Go-based C2 framework — all tokens Claude does not need. | 2 / 3 |
Actionability | Every phase ships concrete, copy-paste-ready commands and configs — install script, systemctl, https/dns/mtls/wg listeners, an NGINX reverse-proxy block, iptables rules, 'generate beacon' invocations, and post-exploitation commands — matching the fully-executable anchor. | 3 / 3 |
Workflow Clarity | The five phases are clearly sequenced and a Validation Criteria checklist exists, but validation is a detached end-of-document list rather than inline checkpoints, and there are no validate→fix→retry feedback loops within the deployment phases. | 2 / 3 |
Progressive Disclosure | Bundle files exist (references/api-reference.md, standards.md, workflows.md; scripts/agent.py, process.py; assets/template.md) yet the body references none of them; all five phases, tables, and signatures sit inline as content that should be split out, matching the 'content that should be separate is inline' anchor. | 2 / 3 |
Total | 9 / 12 Passed |