Content
81%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.
This is a strong operational skill: clearly ordered loops with real validation gates and fallback behavior, concrete commands throughout, and provider detail correctly pushed into a single one-level-deep reference. The main costs are a body long enough that some command catalogs read as inlined reference material, and a handful of steps that defer to external playbooks or a generator script the bundle does not actually provide.
Suggestions
Move the gh provenance-probe and coverage-search command catalogs into a references/ file (like gitcrawl.md) and keep only the decision rules inline, shortening the SKILL.md body.
Provide or at least sketch the ledger generator (or its TSV/Markdown emission format) in a scripts/ or references/ file, so 'regenerate the ledger' is executable rather than delegated to an unspecified local method.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense, imperative, and assumes competence — it teaches a novel ledger workflow rather than concepts Claude already knows, and it delegates provider detail to the gitcrawl reference. It sits at anchor 4 rather than 5 because of minor trimmable padding: the canonical docs path is repeated nearly verbatim in four sections, and cryptic asides like "A ledger family that mixes generated input, manual classification and a summary gets a local check that fails when they drift" restate rules already implied elsewhere. | 4 / 5 |
Actionability | Most guidance is copy-paste ready: concrete gitcrawl/gh/rg commands with explicit JSON field lists, exact artifact paths, an enumerated row schema, and closed sets of owner/status values. It stops short of anchor 5 because a few steps hand-wave: the ledger generator script is never provided ("update the TSV/MD by the smallest reliable local method"), and steps defer to external documents not in the bundle ("the target repo/license gate from `research harvest`", "the Tests rule in `AGENTS.md`"). | 4 / 5 |
Workflow Clarity | The multi-step processes are explicitly sequenced (6-step resume path, 6-step bootstrap path, 8-step --refresh-only, 12-step issue loop) with explicit validation checkpoints ("run that command before marking covered", "Run focused verification", "The refresh is complete only when compact docs artifacts are current or the exact provider blocker is recorded") and genuine feedback loops for a batch operation: provider fallback chain, `needs-repro` when behavior is uncertain, metadata-change re-read without erasing decisions, and unambiguous stop rules. This matches anchor 5, including the batch-operation feedback-loop requirement. | 5 / 5 |
Progressive Disclosure | The body references exactly one bundle file, [gitcrawl](./references/gitcrawl.md), which exists, is one level deep, and is clearly signaled at the point of use — the provider-specific install/archive detail is appropriately split out. It stays at anchor 4 rather than 5 because the SKILL.md is a ~280-line full procedure rather than an overview-plus-references: the gh provenance probes, coverage-search patterns, and the three legacy header-family descriptions are inlined reference material that could live in separate files for easier navigation. | 4 / 5 |
Total | 17 / 20 Passed |