From reverse-engineering-skills
Connect an installed Hopper MCP server safely from Codex or Hermes for document navigation, assembly, pseudocode, strings, procedures, and cross-references with a private profile, read-only allowlist, approvals, and data boundaries.
How this skill is triggered — by the user, by Claude, or both
Slash command
/reverse-engineering-skills:connect-hopper-mcpThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use Hopper's local stdio MCP server only after Hopper and the target document are ready. This workflow connects a local server to an agent client; it does not make Hopper output private to the client model or provider.
Use Hopper's local stdio MCP server only after Hopper and the target document are ready. This workflow connects a local server to an agent client; it does not make Hopper output private to the client model or provider.
Read references/hopper-mcp-profile.md for the data boundary and packaging decision.
Preflight the local application.
Configure a private client profile.
HopperMCPServer command; do not commit machine-local application paths or user configuration.Enable a read/navigation allowlist first.
Gate every disclosure.
Calibrate the tool surface.
Escalate mutations deliberately.
.mcp.json; it is an external local application, not a Socket-distributed server.$use-hopper for interactive document work and $script-hopper-analysis for direct Python SDK work.Return the client profile state, Hopper and server identities, document checkpoint, enabled tool classes, approval/data destination decision, targeted calls, returned evidence, and any mutation escalation. For connection-only work, state preflight only; no document query or mutation performed.
npx claudepluginhub gaelic-ghost/socket --plugin reverse-engineering-skillsCreates, edits, and verifies skills using a test-driven development approach with pressure scenarios and subagents.