Content
63%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 well structured and concise with a clear sequenced procedure and validation checkpoints, but its guidance is more procedural than executable — no concrete inspection commands are given and the HE stages it routes to are undefined. Referenced files (assets icons, session-evidence-contract.md) are missing from the bundle or point outside it.
Suggestions
Add the concrete live-state inspection commands (e.g., the actual branch/PR/CI/Linear check commands to run in step 4) or reference where the HE stage definitions live, so the procedure is executable rather than procedural.
Fix the broken references: create the assets/ directory with the referenced icon files, and correct the session-evidence-contract.md path so it resolves inside the skill bundle instead of escaping via ../../../.
Trim the Philosophy and "Progressive Disclosure Entry" meta sections, which duplicate Failure Modes/Constraints, and fold their unique content into the sections that already carry it.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean, uses terse bullet fragments, and never explains concepts Claude already knows (no definition of Linear, CI, or PR). Minor trimmable material exists: the Philosophy section overlaps Constraints/Failure Modes, and the "Progressive Disclosure Entry" line is meta filler, keeping it below the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Some concrete guidance exists (the 13-outcome taxonomy, "~/.agents/session-collector", "mode:autofix" normalization, "schema_version: 1"), but key details are missing: "re-run or inspect the smallest safe live-state checks" names no actual commands, and the HE stages to route to are never defined in the bundle. Not score 4 because the core inspection loop lacks any executable command or example. | 3 / 5 |
Workflow Clarity | A clear numbered 8-step procedure is backed by an explicit Validation section, a "Fail fast: stop at first failed gate" checkpoint, and Failure Modes covering error paths. Not score 5 because there is no fix-and-revalidate feedback loop after refresh actions are applied. | 4 / 5 |
Progressive Disclosure | Sections are well organized and references are clearly signaled, but scored against the actual bundle the referenced files do not exist: there is no assets/ directory (icon-small.png and icon-large.png links dangle), and the session-evidence-contract.md link escapes the skill bundle via ../../../. Not score 4 because the bundle structure fails the referenced paths; not score 2 because references are clearly signaled, not buried. | 3 / 5 |
Total | 14 / 20 Passed |