Conciseness

The body is dense, terse, and information-rich with zero padding: it assumes Claude's competence and packs only non-obvious behavioral details (exit code 75, lock semantics, `--yes` placement rules, rg-style output format, kind qualifiers). No concept Claude already knows is explained, matching the 'lean and efficient; every token earns its place' anchor rather than the 4 anchor, which requires trimmable over-explanation.

Actionability

Guidance is overwhelmingly executable: concrete copy-paste shell commands (`orbit grep "method:rate limit" src`, `glab orbit --yes grep "..."`, `orbit context mr:42 ...`) with specific flags and behaviors. It falls short of the 5 anchor because of a real gap: the "Choose the command" section introduces `orbit query` with "Write the query in the language below", but no query language appears anywhere in the body or in the single reference file, and the References table link text (`references/cli.md`) does not match the actual path (`references/local/cli.md`).

Workflow Clarity

The core find-then-read workflow (grep → context) is unambiguous with worked examples, command selection is guided by the "Choose the command" section, and the "Fix inaccurate guidance" section has a clear confirm-then-act sequence. Above the 3 anchor because sequences and edge-handling (stale index, async indexing) are explicit; below the 5 anchor because one of the three advertised commands (`orbit query`) has no documented usage, leaving that branch of the decision flow undefined.

Progressive Disclosure

Good structure overall: a focused overview, clearly signaled one-level-deep reference to a real file (`references/local/cli.md`), and a References table for navigation. Not the 5 anchor due to minor gaps: the table's link text says `references/cli.md` while the target is `references/local/cli.md`, and ~50 lines of grep-query semantics are inlined in SKILL.md where the pattern of pointing to a reference file is already established.

Specificity

The description lists multiple concrete actions — "Index and query a local checkout offline", "One call replaces many file reads and text greps", "also searches and reads the project's GitLab" — with concrete query examples ("who calls or extends this, where it is defined, what a file defines"). Coverage is comprehensive, including where the tool works ("working tree and unpushed branches") and where it does not. Not the 4 anchor because there are no meaningful gaps: both local and GitLab capabilities, plus exclusions, are enumerated.

Completeness

Both what and when are answered explicitly and concretely: what — indexing/querying a checkout offline via `orbit`/`glab orbit` replacing greps; when — "Reach for it when a question names a symbol or spans code structure" with example questions, plus an explicit negative trigger ("Not a fit: reading one known file, or counts and relationships across GitLab"). The negative guidance exceeds the 4 anchor's 'when could be more explicit'.

Trigger Term Quality

Natural trigger phrases are well represented: "when a question names a symbol or spans code structure", "who calls or extends this", "where it is defined". Missing a few common synonyms a user might say ("find references/usages", "code search") and any file-extension or command-name triggers beyond the tool's own, which keeps it below the comprehensive 5 anchor.

Distinctiveness Conflict Risk

The tool is named specifically (GitLab Orbit CLI / `orbit` / `glab orbit`), is explicitly distinguished from plain file reads and greps, and explicitly routes GitLab-wide aggregation to the separate `orbit` skill. Clear niche, minimal conflict risk, matching the 5 anchor.