Skill Trust Index / detail
JuliusBrussee/caveman/cavecrew
94
trusted
analyzed 53m ago
WHY THIS SCORE
- -2 · LLM low Safety & Transparency: The skill explicitly documents its output format ('caveman-compressed') and provides clear rules for
- -2 · LLM low Security Awareness: The 'Auto-clarity' section notes that subagents drop to normal English for security warnings and irr
- -2 · LLM low Code Execution Risk: The skill involves file editing (builder) and diff review (reviewer). These are standard, low-risk o
Claims
A decision guide for delegating tasks to three specific subagents (investigator, builder, reviewer) to reduce context window usage by returning compressed, structured output instead of verbose prose.
Actual behavior
The skill acts as a static configuration/manifest. It defines output contracts (formats) and usage rules for subagents. It does not execute code itself but instructs the main agent on when to spawn specific subagents and how to interpret their results. It relies on standard file operations and text processing.
Findings
No non-informational findings.
Attestation
signer 0xB62e1c338a83D3a6621f9127eEa5B000caCfCd01
digest 0x07f94e3b29af9089398aa3f8e172b1bea58c934ba4758fac14dc70a21a42766a
verify: GET /skill/verify?digest=…&signature=… · scheme eip191-sha256
digest 0x07f94e3b29af9089398aa3f8e172b1bea58c934ba4758fac14dc70a21a42766a
verify: GET /skill/verify?digest=…&signature=… · scheme eip191-sha256