Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-organized, concise escalation router with concrete commands and a strong progressive-disclosure structure backed by real reference/script files. The main weakness is that the Tier 1 flagship command references an `engine/` module (and `waf_profiles.yaml`) that is not present in the bundle, leaving the primary tier's core invocation non-executable as written.
Suggestions
Resolve the `engine/` reference: the body states the engine 'lives inside this skill at engine/' and the flagship Tier 1 command is `python3 -m engine "..."`, but no `engine/` directory (and no `waf_profiles.yaml`) ships in the bundle — vendor the module or point the command at an installed package so the default tier runs as written.
Since Tier 1 is the default route for most blocked-URL/article requests, add a one-line fallback using the already-documented primitives (curl_cffi TLS impersonation, Jina Reader, yt-dlp) so the skill stays actionable even before `engine/` is available.
State explicitly how `engine/` is obtained (vendored vs. runtime-installed vs. added by the user) so the command is not a dead reference for whoever loads the skill.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and information-rich — routing diagram, When/Why-first headers, tables, and copy-paste commands — with no padding that explains concepts Claude already knows (no 'what is a WAF/TLS'). It is not the score-2 case because it does not include unnecessary explanation; tokens earn their place. | 3 / 3 |
Actionability | Most commands are concrete and copy-paste ready (yt-dlp with real flags, curl Jina/V2EX endpoints, mcporter, the agent-browser CDP sequence, scripts/extract_cookies.py which exists), but the flagship Tier 1 command `python3 -m engine "..."` and the `waf_profiles.yaml` reference point to a module/file absent from the bundle, so the primary-tier command is incomplete/non-runnable as written. This is the score-2 'some concrete guidance but incomplete; missing key details' rather than score-3 'fully executable'. | 2 / 3 |
Workflow Clarity | A clear tiered sequence (Phase 0 routing diagram → Tier 1 → Tier 1.5 → Tier 2) with explicit escalation checkpoints ('Tier 1/1.5 returned empty or partial -> TIER 2') and a verification note ('reload after injecting' / 'Do not inject cookies without reloading'); the anti-patterns list acts as a checklist. Not score-2 because checkpoints/feedback loops are explicit, and the destructive/batch cap does not apply to a web-access skill. | 3 / 3 |
Progressive Disclosure | A clear overview pointing to well-signaled one-level-deep references, with a 'Reference docs' table mapping each file to when to read it; detailed engine/auth/CDP content is appropriately split into references/*.md and scripts/*.py, all of which resolve in the actual bundle. Not score-2 because structure, signaling, and navigation are all clean. | 3 / 3 |
Total | 11 / 12 Passed |