Content
63%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A largely executable, well-sectioned skill with concrete commands and a sensible step protocol. Weaknesses are duplicated command blocks, no post-execution verification loop, and bundled scripts that are named but never shown how to be called.
Suggestions
Consolidate the duplicated discover/create/execute blocks — keep one canonical CLI Reference and have the protocol steps reference it instead of repeating full commands.
Add a post-run verification step to the protocol (e.g., "6. Verify: `muapi workflow status <run_id>`; if a node failed, fix inputs and re-execute") to create a real feedback loop.
Make the bundled scripts actionable: give their path and flag usage (e.g., `./scripts/generate-workflow.sh --prompt "..." --workflow-id <id>`), or move the CLI/MCP reference tables into a references/ file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is command-driven and mostly lean, but the same commands repeat across sections: `muapi workflow discover` appears in Core Operations, Agent Guided Discovery, Step 4, and CLI Reference; `create`/`execute` also appear three times each. This matches "mostly efficient but includes some... could be tightened" rather than the minor-trim level of 4, since whole duplicate blocks could be collapsed. | 3 / 5 |
Actionability | Nearly everything is copy-paste-ready bash with the `--input "node1.prompt=..."` format shown, and the "CRITICAL RULE" about reading the API Inputs table is concretely actionable. Not a 5: the five bundled scripts in ./scripts/ are named in Core Operations but never invoked — the body shows `muapi workflow create "..."` while `generate-workflow.sh` actually requires `--prompt`, so a reader cannot run the scripts as documented; `pip install /path/to/muapi-cli` also embeds an unresolved placeholder. | 4 / 5 |
Workflow Clarity | The Building-a-Workflow protocol is a clearly numbered 5-step sequence with inspection before execution, and the guided-discovery flow includes confirmation checkpoints and the never-invent-inputs rule. Not a 5: after `execute` there is no explicit verify-recover step (checking `workflow status`, handling a failed node, retrieving outputs is left only to the CLI Reference); not a 3 because the sequence and most checkpoints are explicit. | 4 / 5 |
Progressive Disclosure | Sections are well-organized with clear headers, and bundle scripts do exist, but references are not clearly signaled: scripts are named bare ("generate-workflow.sh") with no path or invocation example, and the full CLI Reference and MCP tool table sit inline in a ~190-line body where a references/ split would fit. This matches "some structure but could be better organized; references present but not clearly signaled". | 3 / 5 |
Total | 14 / 20 Passed |