Content
57%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.
The body delivers strong, command-level actionability and a well-sequenced five-phase testing workflow, and the output template with a full finding example is a highlight. However, it spends tokens re-teaching well-known security concepts, lacks explicit validation/reproduction checkpoints between phases, and fails to reference the api-reference.md and agent.py files that exist in its own bundle.
Suggestions
Remove or drastically trim the 'Key Concepts' table and 'Tools & Systems' section — Claude already knows what certificate pinning, Frida, jadx, and Burp Suite are; keep only mobile-specific nuances it would not infer (e.g., ATS exception patterns).
Add explicit validation checkpoints between workflow steps, e.g., after each phase confirm that each suspected finding is reproducible and record the PoC evidence before moving on, mirroring the evidence requirements already shown in the Output Format.
Link the existing bundle files from the body — e.g., under a 'Automation' or 'API reference' heading, point to [api-reference.md](references/api-reference.md) and scripts/agent.py for automated MASTG checks — so the bundled material is discoverable instead of orphaned.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The workflow steps are lean and command-driven ('jadx -d output/ target.apk', 'sqlite3 ... ".dump"'), but two full sections explain concepts Claude already knows: the Key Concepts table defines 'Certificate Pinning', 'Frida', 'Deep Links', 'App Transport Security (ATS)', and the Tools & Systems section re-describes MobSF, jadx, and Burp Suite. This is 'mostly efficient but includes some unnecessary explanation' — more than the 'minor instances' of anchor 4, but the rest of the body is not padded, so it is above anchor 2. | 3 / 5 |
Actionability | Mostly executable guidance with copy-paste commands at each step ('grep -rn "api_key\|password\|secret\|token\|aws_" output/', 'objection -g "Target App" explore --startup-command "ios sslpinning disable"', 'adb shell am start -n com.target.app/.InternalActivity -e "user_id" "admin"') and a complete PoC in the output template. Minor gaps keep it from anchor 5: several runtime-manipulation steps describe what to do ('hook the detection function to return false', 'Use Frida to replace Objective-C method implementations') without providing the actual Frida scripts, and the Ghidra/Hopper step gives no procedure. | 4 / 5 |
Workflow Clarity | A clear five-step sequence (Static Analysis → Network → Data Storage → Authentication → Runtime Manipulation) with platform-specific sub-steps, but checkpoints are missing or implicit — the only diagnostics are 'If the app fails to connect through the proxy, it may implement certificate pinning' and 'Verify whether the app detects tampering'. No step validates or reproduces findings before reporting. Matches anchor 3 ('sequence present but checkpoints missing or implicit'); not anchor 4 because 'most checkpoints present' is not satisfied by what is written. | 3 / 5 |
Progressive Disclosure | The bundle contains references/api-reference.md and scripts/agent.py, but the body never mentions or links to either — the API reference and automation script are completely disconnected from SKILL.md navigation. The body itself is well-sectioned, so this matches 'some structure... references present but not clearly signaled' (anchor 3); not anchor 4 because an entire bundle file is orphaned rather than having 'minor organization gaps'. | 3 / 5 |
Total | 13 / 20 Passed |