Conciseness

The body is dense and machine-specific throughout — binary resolution order, the client/server RAM behavior, the 'session stop can report a 15s timeout even though the stop landed' caveat, and geometry rules are things Claude cannot know. It never explains general concepts. It falls just short of the 5 anchor because of minor trimmable redundancy: the opening paragraph re-enumerates coverage already in the description, and the 'parse IDs from JSON, never predict/construct them' rule is stated three times (Locate the binary, Discovering state, Safety rules) — deliberate emphasis, but the 4 anchor ('minor instances of over-explanation that could be trimmed') fits better than 5's 'every token earns its place'.

Actionability

Guidance is fully executable and copy-paste ready: real PowerShell invocations with exact flags and values ("& $herdr pane split w1:p1 --direction right --ratio 0.5 --no-focus", "& $herdr wait output w1:p2 --match "test result" --timeout 120000"), exact script usage lines ("pwsh -File scripts/herdr-cleanup.ps1 -DryRun"), and concrete error-recovery commands. This matches the 5 anchor: specific commands cover the common cases, well beyond the 4 anchor's 'minor gaps'.

Workflow Clarity

Sequences are explicit with validation checkpoints exactly where risk lives: cleanup is staged as DryRun inventory → graceful stop → force-kill with an agent-detection backstop that refuses without -Force; agent driving is launch → wait idle → submit → wait done; and error-recovery feedback loops are stated (re-check the process list before concluding 'session stop' failed; 'Inspect before waiting' because wait output can false-match scrollback). Because the destructive/batch operations here DO have explicit validation, the workflow-clarity cap of 3 does not apply; this matches the 5 anchor (clear sequence, explicit validation, feedback loops for error recovery), not the 4 anchor's 'minor validation gaps'.

Progressive Disclosure

Verified against the actual bundle: the ~160-line body is an overview plus the most-used commands, with bulk CLI syntax split to references/cli-reference.md and agent-driving lore split to references/orchestration-playbook.md — both real files, one level deep (the playbook's .md mentions are user task files, not nested skill references). Fiddly Win32 parts are factored into scripts/herdr-window.ps1 and scripts/herdr-cleanup.ps1 with usage documented inline. References are clearly signaled with purpose and read-conditions ("Read it before driving more than one agent"), matching the 5 anchor: clear overview, well-signaled one-level-deep references, easy navigation.

Specificity

The description lists many concrete, specific actions covering the skill's full surface: "open/close/maximize its window, stop the lingering server and reclaim RAM, create/close workspaces, tabs, and panes, manage pane layout (split, balance, zoom, resize, swap, move), launch and stop commands in panes, drive agent CLIs in panes, use the msg bus (send/read/group), and spawn agent teams". This matches the 5 anchor (multiple specific concrete actions, comprehensive coverage) — every major capability in the body is named, exceeding the 4 anchor's "minor gaps in coverage".

Completeness

Both questions are explicitly answered: a detailed 'what' (the enumerated action list) and an explicit 'when' clause ("Use this skill whenever the user mentions... even if they only say 'open herdr', 'kill herdr', 'spawn a team', or 'message the worker pane'") with concrete trigger phrases. This is a direct match to the 5 anchor's exemplar pattern (action list + 'Use when...' with concrete triggers), not merely the 4 anchor's 'when could be more explicit'.

Trigger Term Quality

Trigger coverage is comprehensive and includes natural phrasing users would actually say: "herdr, herdrplus, HerdrPlus panes/tabs/workspaces, spawning or messaging agents in herdr, cleaning up herdr memory/RAM/processes, or orchestrating work across herdr panes — even if they only say 'open herdr', 'kill herdr', 'spawn a team', or 'message the worker pane'". It goes beyond the 5 anchor's synonym/extension coverage by including colloquial example utterances; no common variation is missing.

Distinctiveness Conflict Risk

The niche is clear and triggers are highly specific to HerdrPlus/herdr (fork name, exe name, panes/tabs/workspaces, msg bus), so it would not fire for unrelated skills. However, the body itself acknowledges an upstream herdr skill exists ("Unlike upstream herdr's own skill"), and this description deliberately claims the bare 'herdr' trigger — a minor but real overlap risk with that closely related sibling skill, matching the 4 anchor ('mostly distinct; minor overlap risk with closely related skills') rather than the 5 anchor's 'minimal conflict risk'.