Content
78%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.
The body is a tight, well-structured orchestration guardrail: it gives the main agent an unambiguous three-step job, explicit failure-mode handling, and firm boundaries against doing the subagent's work. Its only weaknesses are minor redundancy between the steps and rules sections and the absence of a concrete launch-command example or report template.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~37-line body is lean with no concept explanations or padding, but the 'Rules for You' section restates the numbered steps (e.g. 'Do NOT try to resume or check on the agent' duplicates step 2's 'do NOT poll, tail'), and 'Trust the agent. It runs in an isolated worktree...' repeats the description. This is 'efficient; minor instances that could be trimmed' rather than every-token-earns-its-place. | 4 / 5 |
Actionability | Guidance is concrete and executable: 'Launch the rulecheck-agent with the focus area (if any): $ARGUMENTS', 'relay its final message verbatim', and an explicit failure protocol ('If the agent hit its context limit, say so... don't go looking for one'). It falls short of fully copy-paste-ready only because no literal launch command or report format example is given. | 4 / 5 |
Workflow Clarity | The launch → wait → report sequence is clearly numbered with explicit failure handling (context limits, agent failure, missing PR URL) and a hard safety stop ('NEVER pick up where the agent left off'). It sits at 'clear sequence with most checkpoints present' rather than 5 because there is no checkpoint for handling partial or ambiguous agent output. | 4 / 5 |
Progressive Disclosure | The skill is under 50 lines, self-contained, needs no external references, and is organized into two clear sections ('Your Job (Main Agent)' and 'Rules for You'), with the one mentioned path (.claude/agent-memory/rulecheck-agent/) as a clear pointer rather than buried content. Per the rubric's simple-skill guideline this earns full marks. | 5 / 5 |
Total | 17 / 20 Passed |