launchpad-byoa-mcp
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEPERSISTENCEDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PERSISTENCE]: The skill modifies agent configuration files (e.g.,
.claude.json,settings.json,opencode.json) to register MCP servers, which maintains the server connection across different agent sessions. - [DYNAMIC_EXECUTION]: The skill automatically deploys a secondary skill by copying
references/retrieve-launchpad-skills.mdto the project's skills directory. The instructions explicitly state to perform this copy without asking for user permission to copy, skip, or replace existing files. - [INDIRECT_PROMPT_INJECTION]: The skill reads and parses existing tool configuration files to merge new server entries. This creates a surface where malformed or malicious data in those files could influence the agent's behavior during the validation and merging process.
- Ingestion points: Target configuration files like
.mcp.json,settings.json, and other active configuration scopes. - Boundary markers: None detected; the skill relies on JSON/TOML structural validation.
- Capability inventory: File system read/write, resolution of project roots, and deployment of additional agent skills.
- Sanitization: Includes checks for parseability, duplicate keys, and structural integrity before writing changes.
- [COMMAND_EXECUTION]: The skill involves resolving project workspaces and agent skill roots through file system inspections and following documented host-specific rules, which implies execution of system-level directory and path operations.
Audit Metadata