Content
85%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 well-crafted conversational-workflow skill: tight, concrete, with explicit validation gates, fallbacks, and error handling throughout. Its single significant defect is that templates/SOUL.template.md is referenced as required reading and is the source of truth for generation, but the file is absent from the bundle — the workflow cannot be executed as written.
Suggestions
Add the missing templates/SOUL.template.md to the bundle, or remove it from the architecture tree and the two 'read before generating' instructions — the generation step depends on it ('Generate the SOUL.md following the template structure exactly', 'Growth section is mandatory and mostly fixed (see template)') and currently points at a file that does not exist.
If the template is intentionally omitted, inline its minimal structure (section headings plus the fixed Growth section text) into SKILL.md so the skill is self-contained and generation is never blocked on a missing file.
Consolidate the field lists that appear three times — the 'Key Extractions' column of the phase table, the Extraction Tracker table, and the per-phase 'Extraction' lines in conversation-guide.md — into the single Extraction Tracker table, and state the pre-read instructions once instead of twice.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is tight and assumes Claude's competence — phase goals, ground rules, and generation rules are delivered via compact tables and terse bullets with no explanation of concepts Claude already knows. It sits at 4 rather than 5 because of repeated redundancy: the 'Key Extractions' column in the phase table, the separate Extraction Tracker table, and the extraction lines in conversation-guide.md restate the same field lists, and the read-both-files instruction appears twice. | 4 / 5 |
Actionability | Fully actionable for an instruction-only skill: explicit pre-read instructions, per-round question limits ("1–3 questions max per round"), a required-field checklist, exact confirmation phrasing, and a concrete tool call — setup_agent(soul="<full SOUL.md content>", description="<one-line description>"). Generation rules even include a concrete good/bad contrast ("write 'argue position, push back, speak truth not comfort' — not 'honest and brave'"). The only caveat is that the template it depends on is absent from the bundle, which is scored under progressive disclosure. | 5 / 5 |
Workflow Clarity | The sequence is explicit with real validation checkpoints and feedback loops: a required-field gate before generation ("You need all required fields before generating"), a hard cap with fallback ("Never exceed 8 — if you're still missing fields, make your best inference and confirm"), iterate-until-confirmed on the generated output, and explicit error handling ("If setup_agent returns an error, report it to the user and do not claim success"). This matches the top anchor's clear sequence with explicit validation and error-recovery loops. | 5 / 5 |
Progressive Disclosure | Structure and signaling are good — a clear overview, one real one-level-deep reference (references/conversation-guide.md, which exists and is well organized) — but the bundle structure does not match the skill: templates/SOUL.template.md is listed in the architecture tree and mandated twice ("Read templates/SOUL.template.md before your first response" / "Read templates/SOUL.template.md if you haven't already"), yet no templates/ directory exists in the bundle. The generation step depends on this missing file ("Generate the SOUL.md following the template structure exactly", "Growth section is mandatory and mostly fixed (see template)") and is immediately qualified as 'if you haven't already' — evidence the author knew this was broken. | 3 / 5 |
Total | 17 / 20 Passed |