Content
82%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 an efficient, well-sequenced operational procedure with concrete commands and a useful report template, and it respects token budget with zero padding. The main gaps are a referenced script that is missing from the bundle and the absence of error-handling steps for common failure cases, which keep actionability, workflow clarity, and progressive disclosure just below top marks.
Suggestions
Ship scripts/setup_experiment.py in the bundle (or replace the Step 1 command with an inline listing method, e.g. 'ls .autoresearch/*/*/results.tsv'), since the referenced path currently does not exist and the command is not executable as written.
Add an error-handling branch to Step 2 for the common failure cases — experiment branch missing, config.cfg or results.tsv absent — so the workflow includes explicit validation checkpoints.
Specify how to derive status in Step 1 (e.g., thresholds on results.tsv age or branch state) instead of the vague 'active/paused/done based on results.tsv age'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a lean, purely operational procedure — every section (usage, list, load context, report, ask next action) contains only instructions, commands, and a concrete output template. No concepts Claude already knows are explained, and nothing is padded, matching the 5 anchor. | 5 / 5 |
Actionability | Mostly executable guidance: concrete commands like 'git checkout autoresearch/{domain}/{name}', 'cat .autoresearch/{domain}/{name}/results.tsv', and 'git log --oneline -20', plus a concrete report template. Not a 5 because Step 1 references '{skill_path}/scripts/setup_experiment.py', which does not exist in the bundle (no scripts/ directory), and 'Show status for each (active/paused/done based on results.tsv age)' leaves the status-derivation logic unspecified — minor gaps in an otherwise copy-paste-ready workflow. | 4 / 5 |
Workflow Clarity | Clear four-step sequence (list if needed → load context → report state → ask next action) with decision checkpoints at steps 1 and 4 (user picks experiment, user picks next action). Not a 5 because there are no error-recovery or validation steps for failure cases (e.g., experiment branch missing, results.tsv absent or corrupt); the operations are not destructive or batch, so the destructive-operation cap does not apply. | 4 / 5 |
Progressive Disclosure | Good structure with clear section headers, appropriately sized inline content for a single-purpose skill, and handoff references to /ar:run and /ar:loop are clearly signaled. Not a 5 because the only bundle reference, scripts/setup_experiment.py, is absent from the actual bundle — a dangling path that breaks navigation — leaving minor organization gaps per the 4 anchor. | 4 / 5 |
Total | 17 / 20 Passed |