ralph

Self-referential loop until task completion with configurable verification reviewer

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Quality

Content

77%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content is highly actionable with a clearly sequenced, well-validated workflow and concrete invocation examples. Its main weaknesses are repetition that inflates token cost and a monolithic structure that should be split across referenced bundle files.

Suggestions

De-duplicate the ai-slop-cleaner skill-vs-agent guidance and the 'do not stop after Step 7 approval' warning so each appears once, trimming token cost.

Move the heavier peripheral sections (company-context interface, parallel session caveats, advanced background-execution rules) into reference files under a docs/ bundle and link to them one level deep.

Either ship the referenced docs (docs/shared/agent-tiers.md, docs/company-context-interface.md, docs/REFERENCE.md) in the bundle or remove the dangling references so navigation is verifiable.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and actionable, but repeats the same guidance multiple times (the ai-slop-cleaner skill-vs-agent distinction appears in Steps 7.5 and Tool_Usage, and the 'do not stop after Step 7 approval' polite-stop warning recurs in Step 7 and Escalation), so it could be tightened rather than scoring as fully lean.	2 / 3
Actionability	It provides concrete, executable guidance throughout — exact Task(subagent_type=..., model=...) calls, Skill("ai-slop-cleaner") invocation, omc ask codex commands, /oh-my-claudecode:cancel, specific file paths, tier mappings, and copy-paste-ready good/bad examples.	3 / 3
Workflow Clarity	Steps 1–9 are clearly sequenced with explicit validation checkpoints (per-criterion verification, tiered reviewer approval, mandatory deslop, regression re-verification) and feedback loops (rejection→fix→re-verify, regression failure→rollback→rerun), plus a Final_Checklist.	3 / 3
Progressive Disclosure	The SKILL.md is a ~250-line monolithic file with PRD mode, execution policy, steps, tool usage, examples, escalation, caveats, and advanced rules all inline rather than split into separate files, and it references docs (agent-tiers.md, company-context-interface.md, REFERENCE.md) that are not present in any bundle directory to verify.	2 / 3
	Total	10 / 12 Passed

Description

35%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description conveys the core concept of a persistence loop with verification but omits any explicit usage triggers and relies on internal jargon rather than natural user language. It is identifiable but not self-activating.

Suggestions

Add an explicit 'Use when...' clause with natural trigger terms users would actually say (e.g. 'ralph', 'don't stop', 'keep going until done', 'must complete').

Replace jargon like 'self-referential loop' and 'configurable verification reviewer' with concrete user-facing phrasing (e.g. 'Keeps working on a task until every user story passes reviewer verification, then cleans up.').

Add a distinguishing phrase to separate it from sibling skills like autopilot, ultrawork, or the native /goal loop.

Dimension	Reasoning	Score
Specificity	Names the domain ('Self-referential loop until task completion') and one mechanism ('configurable verification reviewer') but does not list multiple concrete actions, fitting the 'names domain and some actions' anchor rather than the vague level-1 or the multi-action level-3.	2 / 3
Completeness	It states what the skill does (persistence loop with verification) but provides no 'Use when...' clause or explicit trigger guidance, so 'when' is entirely missing — capping completeness at 2 per the rubric guideline.	2 / 3
Trigger Term Quality	The description contains only internal jargon ('self-referential loop', 'configurable verification reviewer') with no natural keywords a user would actually say; the genuine triggers (e.g. 'ralph', 'don't stop', 'must complete') appear only in the body, not the description.	1 / 3
Distinctiveness Conflict Risk	'loop until task completion with configurable verification reviewer' is somewhat specific but overlaps with sibling autonomy skills (autopilot, ultrawork, /goal noted in the body) and lacks distinct triggers to prevent misfiring.	2 / 3
	Total	7 / 12 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: Yeachan-Heo/oh-my-claudecode
Commit: 6c59219

Reviewed: about 7 hours ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.