Conciseness

The body is lean with no padding of concepts Claude already knows, but the opening line — 'This is the v1.8+ canonical loop skill name. It supersedes `autonomous-loops` while keeping compatibility for one release' — is time-sensitive version metadata that is not placed in a deprecated/old-patterns section, which the guidelines say should penalize conciseness. Not level 2, since it is one line rather than 'several unnecessary explanations or padded sections'.

Actionability

Guidance is high-level hints without executable specifics: 'Recommended production stack' only names other skills (`ralphinho-rfc-pipeline`, `plankton-code-quality`, `/quality-gate`, `eval-harness`, `nanoclaw-repl`), and recovery items like 'freeze loop', 'reduce scope to failing unit', and 'run `/harness-audit`' give no commands, parameters, or criteria. The selection flowchart is the only quasi-decision procedure, and even it lacks instructions for executing the chosen loop — matching level 2's 'high-level hints but missing the specific steps to execute'.

Workflow Clarity

Sequences are listed (combined pattern stack 1-4; recovery ordered freeze → audit → reduce scope → replay) but validation checkpoints are absent or implicit. Notably, a skill whose description centers on 'gate on evals' never shows how to check a gate or what acceptance criteria look like, matching level 3's 'sequence present but checkpoints missing or implicit'.

Progressive Disclosure

At under 50 lines with well-organized sections (Loop Selection Flow, Combined Pattern, Failure Modes, Recovery) and no bundle files to disclose, the simple-skill rule would allow a 5. It drops one level because the body references six named skills and commands that are not bundle files and are not clearly signaled as external cross-references — matching level 4's 'references mostly clear; minor organization gaps'.

Specificity

Names the domain ('continuous autonomous agent loops') and its constituent elements ('quality gates, evals, and recovery controls'), but these are themes rather than concrete actions — closer to 'names domain and 1-2 concrete actions, but not comprehensive' than to the 'several specific actions' of level 4.

Completeness

Both parts are present: a clear what ('Patterns for continuous autonomous agent loops with quality gates, evals, and recovery controls') and an explicit when ('Use when running an agent loop that must self-check, gate on evals, and recover from failures'). The when restates the what's vocabulary rather than adding distinct concrete trigger phrases, matching level 4's 'when could be more explicit or specific' rather than level 5.

Trigger Term Quality

The 'when' clause supplies natural phrases users would say — 'running an agent loop', 'self-check', 'gate on evals', 'recover from failures' — plus 'autonomous loops' and 'continuous' in the what-clause. Missing common synonyms/variations a user might type (e.g. 'background loop', 'keep running until done'), so it falls just short of level 5's comprehensive synonym coverage.

Distinctiveness Conflict Risk

The niche is fairly clear — self-checking agent loops with eval gates — and unlikely to trigger for unrelated skills. There is minor overlap risk with generic loop/iteration or CI/PR skills whose users would also say 'agent loop', keeping it below level 5.