Content
88%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 strong, highly actionable skill body: executable commands, explicit gates and defaults, resumable checkpoints, and fail-closed validation for a batch operation. Its main weaknesses are mild redundancy (terminal-precedence stated twice, When-to-Use echoing the description) and a somewhat monolithic main file where historical/operational detail could move to references.
Suggestions
Collapse the 'Migration from the old precedence/fallback behavior' blockquote into the Step 1 precedence line — one sentence stating the order (--terminal-path > config.terminal_path > MT5_TERMINAL_PATH, hard error if missing) already conveys it without the migration history.
Trim the 'When to Use' section to a single line (e.g., 'Use when the user asks to batch-test or screen MT5 EAs, optimize parameters, or resume a run') since the frontmatter description already enumerates the same triggers.
Move the Round-2 reference thresholds and the detailed Key Principles (e.g., OS-lock/.blocked handling, build-verification note) into references/mt5-cli-reference.md or a short operations reference, keeping SKILL.md as a lean overview pointing to them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with skill-specific, non-inferable knowledge (build 6061 XML quirks, relative Report= names, gate thresholds, terminal-precedence rules, the .blocked marker) and explains nothing Claude already knows, which places it above the 'mostly efficient' level-3 anchor. However, the two long blockquotes on terminal selection and the 'Migration from the old precedence/fallback behavior' paragraph partially restate the precedence chain already given in Step 1, and the 'When to Use' list overlaps the frontmatter description — minor padding that keeps it below 'every token earns its place'. | 4 / 5 |
Actionability | Guidance is fully executable: copy-paste-ready commands for dry-run, execution, resume, and the dashboard (`python3 skills/mt5-robot-tester/scripts/mt5_batch_tester.py --config my_config.json ...`), a named config template to copy (`assets/pipeline_config.template.json`), exact gates with defaults (`≥5 symbols profitable`, `best symbol ≥3× deposit`, `profit ≥4× deposit`, `worst DD ≤12%`, `LR Correlation ≥0.80`), and a precise optimization recipe (`[V×0.5, V×1.5]` step `V×0.05`). This matches the anchor 'Fully executable; copy-paste ready code or commands; specific examples cover the common cases'. | 5 / 5 |
Workflow Clarity | The pipeline is sequenced as explicit Steps 1–5 (configure → dry-run → run → resume → read results) with round-by-round gates, and — critical for a batch skill — validation is explicit throughout: the offline `--dry-run` verification step, the fail-closed principle ('incomplete, timed-out, stale, or unparsable reports never reject, promote, or move a candidate'), fingerprint-checked `--resume`, and the `.blocked` marker procedure. This matches the anchor 'Clear sequence with explicit validation steps; feedback loops for error recovery', including the batch-operation validation requirement. | 5 / 5 |
Progressive Disclosure | The Resources section clearly signals one-level-deep bundle files that all exist on disk (`scripts/mt5_batch_tester.py`, `scripts/parse_mt5_optimization.py`, `scripts/parse_mt5_report.py`, `scripts/mt5_learnings.py`, `scripts/mt5_common.py`, `references/mt5-cli-reference.md`, `assets/pipeline_config.template.json`), and detailed tester key/enum material is correctly deferred to the reference. However, ~205 lines of operational detail — round-gate specifics, the terminal-selection safety note, the migration note, and nine Key Principles — are inlined in SKILL.md where some (migration history, Round-2 reference thresholds) would fit better in a reference file, so structure is 'good' rather than the ideal 'content appropriately split'. | 4 / 5 |
Total | 18 / 20 Passed |