Content
96%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.
An exemplary router-style body: lean, fully executable commands verified against the bundled CLI, explicit sequencing with validation gates, stop conditions, and feedback loops. The only weakness is navigation depth — action references chain one level deeper via shared-*.md files, and the four docs.*.md bundle files are orphaned with no entry point pointing to them.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 49-line body is pure procedural instruction ("Treat the first word as `$ACTION`, remainder as `$REST`", "Run the init gate, except for `explore`") with zero explanation of concepts Claude already knows. The densest passages ("rewrite each affected section as it would read if the current target had been known from the start") are skill-specific policy, not padding, so there is nothing to trim — matching the anchor 'lean and efficient; assumes Claude's competence' rather than anchor 4's 'minor instances of over-explanation'. | 5 / 5 |
Actionability | Every command is copy-paste ready and verified against the bundle: `node <skill-dir>/scripts/cli.mjs init --status`, `status "<name>" --apply`, and `list` all exist in scripts/cli.mjs and its actions, and the dispatch logic is fully concrete ("If uninitialized and `$ACTION` is not `init`, suggest `/esdd init` and stop"). This matches 'fully executable; copy-paste ready commands; specific examples cover the common cases' — the no-input, invalid-input, valid-action, and follow-up cases are all handled. | 5 / 5 |
Workflow Clarity | Steps 1–6 give a clear sequence with an explicit validation checkpoint (the Init Gate's if/then/stop conditions) and genuine feedback loops: "Check tasks only when implementation and relevant verification satisfy their outcomes; uncheck unmet ones. Run relevant tests and build/type checks after code corrections", plus stop-on-blocked revision. It fits anchor 5 ('explicit validation steps; feedback loops for error recovery') rather than anchor 4 because checkpoints are present at every decision point, not just 'most'. | 5 / 5 |
Progressive Disclosure | The body is a lean overview dispatching to ten action references, all of which are real files, plus shared-revision.md and the CLI script — good structure matching anchor 4. It falls short of anchor 5 ('well-signaled one-level-deep references') because action files chain to a second markdown level (e.g. continue.md → shared-revision.md, shared-plan-loop.md) and four bundle files (docs.commands.md, docs.workflows.md, docs.artifacts.md, docs.configuration.md, ~23KB total) are referenced from nowhere — not SKILL.md, any reference file, or the scripts — leaving that content unreachable via navigation. | 4 / 5 |
Total | 19 / 20 Passed |