Content
73%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 well-sequenced, validation-heavy workflow for genuinely risky operations, with concrete column specs, statuses, and paths that make it highly actionable as an instruction skill. Its main cost is token efficiency: governance rules are restated across nearly every section, and referenced-file content is duplicated inline rather than trusted to the references. The bundle also ships no scripts or references despite naming connector scripts and external protocol docs.
Suggestions
State the never-write-canonical-registry / propose-via-registry-events.py rule once (e.g., in the Skill Contract Writes bullet) and drop its repetition in the scope guard, step 6, Save Results, and Reference Materials.
Consolidate the one-intent/one-receipt-per-irreversible-action rule (currently spread across steps 3, 4, 5, and 7) into a single stated principle, referencing action-control.md for the details instead of restating its semantics inline.
Either ship the referenced connector scripts (scripts/connectors/hn.py, producthunt.py, appstore.py, gdelt.py) in the bundle or trim the Data Sources section to the CONNECTORS.md pointer so named-but-absent files don't mislead.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body never explains concepts Claude already knows, but the same governance rules are restated many times — the propose-via-registry-events.py restriction appears in the scope guard, the Writes section, step 6, Save Results, and Reference Materials, and the one-intent/one-receipt rule recurs across steps 3, 4, 5, and 7 — so it could be tightened considerably. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is concrete: an explicit runbook column spec, receipt status vocabulary (succeeded | partial | failed | unknown), exact memory paths, and named connector scripts. It falls short of fully executable because key semantics (action-control, state-model proposal resolution) live in external files not present in this bundle. | 4 / 5 |
Workflow Clarity | Eight clearly sequenced steps with explicit validation checkpoints for every irreversible action — a hard gate that stops with NEEDS_INPUT, pre-declared kill-criteria requirements, per-action receipts, rollback as a separately receipted action, and open-lane joins on missing receipts — plus explicit Done-when criteria. | 5 / 5 |
Progressive Disclosure | Clear section structure with a Reference Materials section that signals nine one-level-deep references, each with a one-line purpose. However, content owned by those referenced files (state-model clauses, action-control semantics) is repeatedly summarized inline, and no bundle files (references/, scripts/, assets/) ship with the skill to back the connector paths named in Data Sources. | 4 / 5 |
Total | 16 / 20 Passed |