ruflo-goals

Reviewed at commit 13af5f1

Caution: Pre-approves Bash, Write and WebFetch in its skills, and its /goals script runs unpinned @claude-flow/cli@latest through npx -y when no ruflo CLI is installed.

Sends data to Web search and the public web pages Claude fetches, only when you run the research or dossier skills

Long-horizon goal planning, deep research orchestration, and adaptive replanning using GOAP algorithms

Made by
ruvnet
What we checked

Adds goal planning, deep research and dossier skills that store plans and findings in ruflo memory, plus a guard that refuses memory and task writes holding a secret or an SSN-shaped number. The research and dossier skills search and fetch web pages; a helper script can run an npm package through npx.

Review details

Static review of the source at the pinned commit; not run. Delta review of the changes since 87d8687; the rest carries over from the review at that commit. The changes are a version bump (0.4.5 to 0.4.6), a repository field in plugin.json, a smoke.sh test script update and one line in the research-synthesize skill that now tells Claude to synthesize findings itself instead of calling the agentdb_context-synthesize tool (no change to its allowed-tools). The hook code and scripts are unchanged. Hook module: session.start (reads the project root, registers /goals-mod, writes .claude-flow/goals-mod/status.json), tool.call (only memory_store, agentdb_hierarchical-store, agentdb_pattern-store, agentdb_batch, hooks_intelligence_pattern-store and task_create; for namespaces horizon, research, goap, dossier, goals it denies input holding a secret shape, and for dossier and research namespaces an SSN-shaped number; deny only, never allows or rewrites) and command.run for /goals-mod (status, scan <text>, horizons). The horizons verb calls the connected memory_list tool read-only through $.mcp.call and shows the first 1200 characters. The hook code has no network, process or model calls. The rest of the plugin is prompts: skills deep-research and dossier-collect pre-approve Bash, WebSearch, WebFetch, Read and Write (plus the ruflo MCP memory and agentdb tools); goal-plan, horizon-track and research-synthesize pre-approve Bash, Read, Write and ruflo MCP tools. Research is capped by a USD budget marker file .claude-flow/research-active.json; the prompt text tells Claude to screen fetched pages with aidefence_scan, treat them as data, fetch only public http(s) URLs and write only the dossier path in the project. The dossier skill builds profiles of usernames and other seeds from public sources, as disclosed. scripts/research-list.mjs (run by /goals through node) lists research records by spawning the ruflo CLI without a shell: a built local CLI, else ruflo or claude-flow found on an absolute PATH entry, else npx -y @claude-flow/cli@latest (via lib/ruflo-cli.mjs); the agents also suggest npx @claude-flow/cli@latest commands via Bash. The skills themselves do not tell Claude to run npx. Not reviewed: the @claude-flow/cli package, the ruflo-core MCP server, and the test scripts and smoke.sh (not loaded by the hooks).

What it uses
  • Reads your prompts: no
  • Sees tool calls: yes
  • Blocks or changes tool calls: yes
  • Reads or writes files: yes
  • Runs programs on your machine: yes
  • Calls the model (uses tokens): no
  • Sends data over the network: yes
  • Stores data: yes
  • Shows something on screen: yes
Hooked events
Made this mod? Add a badge to your README

modsforClaude badge for ruflo-goals

Paste this into your README. The badge says reviewed once the current version has been reviewed, and links readers to this page.

[![Listed on modsforClaude](https://modsforclaude.com/badge/ruflo-goals.svg)](https://modsforclaude.com/mods/ruflo-goals/)

← All mods