launchdarkly-flag-cleanup
Pass
Audited by Gen Agent Trust Hub on Apr 17, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is authored by the official vendor (LaunchDarkly) and integrates with their designated MCP server and tools.
- [PROMPT_INJECTION]: Instructions in
SKILL.mdfollow a strict, multi-step workflow focused on safety and verification. There are no attempts to override agent behavior or bypass system constraints; instead, it explicitly instructs the agent to wait for user confirmation before proceeding with modifications. - [DATA_EXFILTRATION]: No patterns for exfiltrating sensitive local data (like SSH keys or AWS credentials) were found. The skill searches for feature flag keys in the codebase to facilitate code removal, which is its primary purpose.
- [REMOTE_CODE_EXECUTION]: The skill relies on the
@launchdarkly/mcp-server, a vendor-owned resource. It does not perform arbitrary remote code execution or download scripts from untrusted sources. - [COMMAND_EXECUTION]: The skill instructs the agent to run project-specific build and lint commands as a verification step. This is standard practice for a developer-focused tool and is confined to the project's own environment.
- [DATA_EXPOSURE]: The skill processes data from the local codebase and the LaunchDarkly API. While this presents a surface for indirect prompt injection (e.g., malicious instructions hidden in code comments), the skill's requirement for a human review checkpoint (Step 4) serves as a significant mitigation.
Audit Metadata