ruflo-workflows
Reviewed
at commit 704330e
Caution: Its skills pre-approve Bash and have Claude run the unpinned npm package @claude-flow/cli@latest through npx, so each new version runs without asking.
Workflow automation across two surfaces: the 10 workflow_* MCP tools (create/run/execute/status/list/pause/resume/cancel/delete/template) with full…
0 upvotes
Log in to upvote or save- Made by
- ruvnet
- Pinned commit
-
704330e488992d65791e76fde59b36af049aa3a3 - What we checked
-
Refuses workflow definitions, runs and GAIA run records that hold something shaped like a key, token or password, and refuses a workflow delete that names no single id. Adds a /wf-mod command and a small counter file in your project. The bundled commands and skills also drive GAIA benchmark runs that spend tokens on your own API keys.
Review details
Static review of the source at the pinned commit; not run. Delta review of the changes since 828a8d7; the rest carries over from the review at that commit. The module registers session.start (project root, registers /wf-mod, writes the status file), tool.call and command.run. tool.call only looks at workflow_create/run/execute/template, workflow_delete, and memory writes (memory_store, agentdb_hierarchical-store, agentdb_pattern-store) in gaia or workflow namespaces: it denies a delete whose workflowId is empty or a wildcard such as all or *, and denies calls whose flattened input matches the regex and entropy secret checks (names only, input too large to screen is refused); everything else passes unchanged (tighten-only; guard option default on). Runtime calls: $.session.root, $.command.register, $.clock.now, $.fs.write (only .claude-flow/wf-mod/status.json in the project, rewritten on each watched call), $.tool.list for /wf-mod managed. No network, no process. scripts/gaia-audit.mjs, run by the gaia-submit command, is a deterministic local audit: it reads the results, trajectory, metadata files and the gaia-bench runner sources, may read ~/.cache/huggingface and ~/.cache/ruflo/gaia/dataset paths, and writes only the --out report; it contains no network calls, eval or child process execution of its own. The prompt layer is larger: the skills gaia-architecture-comparison, gaia-debugging, gaia-submission, workflow-create and workflow-run pre-approve Bash and the commands, skills and agents run npx @claude-flow/cli@latest for memory and hooks commands; gaia-run and gaia-debugging run node $(npm root -g)/@claude-flow/cli/bin/cli.js gaia-bench run or node v3/@claude-flow/cli/bin/cli.js from the opened project, which spends Anthropic tokens (the command asks for confirmation above about $5) and uses ANTHROPIC_API_KEY, HF_TOKEN and optional Google keys from the environment to download the GAIA dataset from Hugging Face; gaia-validate runs npx tsc --noEmit and a smoke question; the submit flow builds a zip and only tells the user where to upload it. workflow-create can write .claude/workflows/*.js and run node --check on a temp file. The @claude-flow/cli benchmark runner, ruflo-core MCP server, the other repo scripts (witness/regen.mjs) and scripts/smoke*.sh were not reviewed. Changes in this version: version bump to 0.7.2, a repository field, the smoke script version, and a new gaia-debugging step that suggests re-running the same gaia-bench command with --enable-critic or --decompose (more model calls on your own keys; no new hosts, installs or allowed-tools); no code or hook changes.
- 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): yes
- Sends data over the network: no
- Stores data: yes
- Shows something on screen: yes
Workflow automation across two surfaces: the 10 workflow_* MCP tools (create/run/execute/status/list/pause/resume/cancel/delete/template) with full state-machine lifecycle (created → running ↔ paused → completed/cancelled), and native Claude Code Workflow JS orchestration (.claude/workflows/*.js — agent/parallel/pipeline/phase fan-out). Includes GAIA benchmark component for Princeton HAL leaderboard submissions.
Made this mod? Add a badge to your README
Paste this into your README. The badge says reviewed once the current version has been reviewed, and links readers to this page.
[](https://modsforclaude.com/mods/ruflo-workflows/)