Content
88%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.
A strong operational skill: fully executable commands, explicit pre-flight checks, and real feedback loops for every failure mode (denial, timeout, no-setup, unlinked contact). Its weaknesses are minor — some redundant guidance between the flag section and earlier sections, and reference material inlined where progressive disclosure into a bundle file would keep the body leaner.
Suggestions
Trim the `--to`/`--via` flag bullet down to flag semantics only — the contacts lookup and `human invite` workflow it repeats are already covered in the 'Who can I reach' section.
Move the full flags reference (card chrome, Slack-only --icon, --ttl, approve/choose-only flags) and the invite/linking mechanics into a references/ file, keeping SKILL.md to the when-to-use decision flow, setup check, and the four commands.
Condense the quoted keyless-demo error output to its actionable core ('run: human login') instead of reproducing the full multi-line message.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with non-obvious operational facts (exit codes, invite-link semantics, env vars) and assumes Claude's competence, but the `--to` flag bullet ("Find them with `human contacts --json` first; link someone new with `human invite alice --name 'Alice Chen'`") repeats the dedicated contacts and invite sections, and the full multi-line keyless-demo error output is quoted verbatim when its actionable core ('run: human login') would do. | 4 / 5 |
Actionability | Every section gives copy-paste-ready commands covering the common cases: the four commands with realistic arguments ("human approve 'Delete 342 rows flagged as stale from the orders table?'"), a complete exit-code branching bash script, `human contacts --json`, `human invite alice --name 'Alice Chen'`, and flag combinations (--icon, --subtitle, --extra-action). | 5 / 5 |
Workflow Clarity | The flow has explicit pre-checks ("Before you rely on it: check it's set up", "look before you ask") and genuine error-recovery loops: exit 11 resumes via "human wait <id>", setup failure is surfaced rather than looped, and delivery failure names the exact `human invite` command to run. The destructive-ops cap does not apply since the skill obtains approvals rather than performing destructive operations itself. | 5 / 5 |
Progressive Disclosure | Well-sectioned single file with clear headers and no nested references, but at ~215 lines it exceeds the 50-line simple-skill threshold and inlines reference-grade material (the full flags list including Slack-only --icon, card chrome, and --ttl; invite/linking mechanics; quoted error texts) that could sit in a references/ file while SKILL.md keeps the decision flow and four commands. | 4 / 5 |
Total | 18 / 20 Passed |