Conciseness

The body is dominated by executable commands, expected-output blocks, and two tight tables — largely lean. Minor over-explanation remains: the intro line "an MCP server exposing 1000+ scientific tools plus 120+ research skills — all auto-configured" duplicates the frontmatter description, and the sample questions ("What are the top mutated genes in breast cancer?" / "Research the drug metformin.") are demonstration padding. This matches anchor 4 ('efficient; minor instances of over-explanation that could be trimmed') better than 5, since not every token earns its place, and is far above the 'mostly efficient but some unnecessary explanation' level 3.

Actionability

Every step is a concrete, copy-paste-ready command: `uv --version`, `agy plugin install /path/to/ToolUniverse/plugin`, the full `git clone https://github.com/mims-harvard/ToolUniverse.git` sequence, `agy plugin list` with expected JSON output, `uv cache clean tooluniverse`, and `uv python install 3.12`. The common cases (fresh install, verify, update, troubleshoot) are each covered by an executable example — anchor 5; the only placeholder (`/path/to/...`) is inherent to user-specific paths and Option 2 removes even that.

Workflow Clarity

The sequence is clear (Prerequisites → Install → Restart → Verify) with an explicit validation checkpoint — "Expect output showing `tooluniverse` under `imports`" gives the exact success criterion — and the Troubleshooting table provides symptom→fix feedback loops for error recovery (`uvx: command not found`, "MCP server won't start", "tools missing"). This matches anchor 5; no destructive or batch operations exist that would cap the score, and validation is explicit rather than implicit, ruling out 4.

Progressive Disclosure

Sections are well organized and self-contained with no nested or dead references (no references/, scripts/, or assets/ exist), so navigation is easy. However, the body runs ~100 lines — above the sub-50-line simple-skill exception — and detail such as the "What you get" component table, the API-key section, and the expanded troubleshooting entries could live in one-level-deep reference files to keep SKILL.md a leaner overview. That places it at anchor 4 ('good structure; most content appropriately placed; minor organization gaps') rather than 5.

Specificity

"Install, set up, verify, update, pin, uninstall, or troubleshoot the ToolUniverse plugin" lists multiple specific concrete actions, and the description further names concrete artifacts like "the exact CLI name (agy)", "the 'agy plugin install <path>' flow", "the uvx tooluniverse MCP server", and "the API-key env vars". This comprehensively matches the anchor-5 example profile; there are no coverage gaps that would justify a 4, and it is far above the vague 'names the domain but actions are minimal' level 2.

Completeness

It explicitly answers 'what' ("Install, set up, verify, update, pin, uninstall, or troubleshoot the ToolUniverse plugin on Google Antigravity") and 'when' ("Use it whenever someone wants to get ToolUniverse ... working on Antigravity, says ... won't load, hits a uvx or MCP-server startup error, asks how Antigravity updates it, wants to pin or remove it, or finds it running an old tool version"), with concrete trigger phrases — the anchor-5 example shape. The 'when' is explicit, not merely implied, ruling out 4.

Trigger Term Quality

The description covers natural phrases a user would actually say: "says the Antigravity plugin or its tools/skills won't load, hits a uvx or MCP-server startup error, asks how Antigravity updates it, wants to pin or remove it, or finds it running an old tool version", plus synonyms like "the 1000+ scientific tools" / "the harvard tools" and name variants "AGY / Antigravity IDE / Antigravity 2.0". This matches the anchor-5 pattern of comprehensive natural-term coverage including synonyms, well above the 'a few natural terms missing' level 4.

Distinctiveness Conflict Risk

The skill is bound to a specific niche (ToolUniverse on Google Antigravity) and explicitly disambiguates sibling skills by name: "Not for the Codex plugin (use tooluniverse-codex-plugin) or Claude Code plugin (use tooluniverse-claude-code-plugin), for running research with the tools, or for authoring new tools or skills." This clear niche with distinct triggers and explicit conflict avoidance matches anchor 5; minor overlap with those sibling skills is explicitly resolved, ruling out 4.