Content
77%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable safety skill with clear confirmation-gated workflows, weakened only by length that could be trimmed and a monolithic structure with a dangling external reference. The guidance itself is excellent.
Suggestions
Tighten rule 5 (public auth): the four-step procedure plus the 'when in doubt' default can be condensed without losing the hard-pause semantics.
Resolve or remove the 'secrets-and-integrations playbook' reference, or create that file under references/ so the inline citation points to a real bundle asset.
Consider splitting the 'Choosing an approval type' section into a references/ file and keeping SKILL.md as a concise overview, which would improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and concrete (no padding explaining concepts Claude already knows), but at ~260 lines some sections — notably rule 5 on public auth — repeat and elaborate beyond what the procedure needs, so it could be tightened. | 2 / 3 |
Actionability | Provides exact executable procedures with named tools ('call posthog__agent-applications-revisions-promote-create'), literal confirmation phrases ('promote', 'ship', 'go'), and concrete bad-pattern quotes to refuse — fully actionable instruction-only guidance. | 3 / 3 |
Workflow Clarity | Destructive flows (promote, public auth, destroy) are sequenced with explicit human-confirmation checkpoints ('State → Ask → Wait → Then call'), which serve as the required validation gate for risky operations. | 3 / 3 |
Progressive Disclosure | Well-organized into clear sections, but it is a single monolithic file with all content inline and one unresolvable reference (the 'secrets-and-integrations' playbook) that has no corresponding bundle file in references/. | 2 / 3 |
Total | 10 / 12 Passed |