The Agent Skills Directory

[SAFE]: The skill's primary function is to perform static analysis and guided refactoring within the local repository environment. It focuses on identifying LLM call sites and adding descriptive workflow labels to HTTP headers or SDK configurations. The skill does not introduce external dependencies, exfiltrate sensitive data, or attempt to acquire elevated privileges. All external resource references, such as the Caveman SDK and gateway headers, are consistent with the skill's stated purpose and the author's context.- [SAFE]: While the skill possesses the capability to modify repository files and execute local commands (e.g., tests or dev scripts), it implements a mandatory human-in-the-loop safety pattern. The instructions explicitly require the agent to propose a summary table and wait for user approval before applying any changes. This significantly mitigates risks associated with potential indirect prompt injection from repository content. Analysis of the capability surface (Category 8) confirms that ingestion of untrusted repository code is balanced by clear boundary markers and operational transparency.

caveman-discover