Conciseness

The body assumes Claude's competence throughout (no basic explanations) and nearly every line is Herdr-specific knowledge Claude cannot know, e.g. "Treat every ID as an opaque string" and the enumerated read sources. It is not a 5 because sections could be tightened: the idle/done semantics span three dense paragraphs of edge cases ("If the foreground client is explicitly unfocused, completion can become done even in the active tab"), and points are repeated ("Inspect before waiting" appears at lines 150 and 192; the split command is shown twice). It is clearly above anchor 3 since there is no generic padding.

Actionability

Quotes: "test "${HERDR_ENV:-}" = 1", "herdr pane split --current --direction right --no-focus", "herdr wait agent-status <returned-pane-id> --status done --timeout 120000", "herdr pane read <returned-pane-id> --source recent-unwrapped --lines 120". Every workflow step is a copy-paste-ready executable command with real flags, timeouts, and JSON field names ("Read result.pane.pane.pane_id"), plus concrete error-handling decision rules. This matches anchor 5; anchor 4 would imply minor gaps in executable coverage, which are not present.

Workflow Clarity

Sequences are explicit with validation checkpoints and feedback loops: env gate ("If the check fails... stop"), "Inspect the pane after launch. If agent_status is not yet idle, wait for the idle transition", "If a wait times out, inspect herdr pane get... and pane read before deciding what to do", and "Re-read create, split, move, list, or get responses after mutations". The pre-flight guard against accidental mutation ("Do not probe a mutating nested command by omitting arguments") strengthens the safety sequence. This matches anchor 5's explicit validation with error-recovery loops; anchor 4 would require missing checkpoints.

Progressive Disclosure

The body is well-sectioned (Learn the current CLI, IDs and current context, Start agents interactively, Run an ordinary command, Safety rules) and wisely delegates CLI syntax to runtime discovery via "herdr --help" instead of inlining version-sensitive help text, which is appropriate one-level-late disclosure. However, the skill exceeds the simple-skill threshold at ~190 lines with no reference layer: the agent-status state machine and the read-source catalog are reference-grade detail living inline, and the bundle's scripts/ and assets/ are never referenced from the body. This lands at anchor 4 (good structure, minor organization gaps) rather than 5.

Specificity

Quotes: "Control Herdr, a terminal multiplexer for coding agents" and "inspect or control panes, tabs, workspaces, terminals, commands, or communication with another agent" — several concrete actions over enumerated resources. It sits at anchor 4 rather than 5 because the verb coverage is umbrella-like (control/inspect) and doesn't enumerate distinct operations (split, read, wait, rename) found in the body; it is clearly above anchor 3, which expects only 1-2 concrete actions.

Completeness

What: "Control Herdr, a terminal multiplexer for coding agents." When: "Use only when the user explicitly mentions Herdr or asks to use Herdr..." plus an explicit exclusion clause and the requirement "Requires HERDR_ENV=1". Both what and when are answered explicitly with concrete trigger phrases, matching anchor 5 exactly; anchor 4 would require the 'when' to be less explicit, which it is not.

Trigger Term Quality

Quotes: "explicitly mentions Herdr", "panes, tabs, workspaces, terminals, commands, or communication with another agent", plus the anti-triggers "background terminal, delegation, or parallel work". Since the natural user trigger is the product name itself, coverage is comprehensive including synonym phrases for what the tool manages; nothing natural is missing, so anchor 5 fits better than 4.

Distinctiveness Conflict Risk

Quotes: "Use only when the user explicitly mentions Herdr" and "Do not use merely because a task could benefit from a background terminal, delegation, or parallel work." The proper-noun gate plus explicit anti-triggers give a clear niche with minimal conflict risk, matching anchor 5 rather than anchor 4's 'minor overlap risk'.