aif-loop

Run a strict multi-iteration Reflex Loop with phases (PLAN, PRODUCE||PREPARE, EVALUATE, CRITIQUE, REFINE) to improve an artifact until quality gates pass or iteration limits are reached. Use when user asks for iterative refinement, quality-gated generation, or "generate -> critique -> refine" loops.

Quality

82%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

81%Weight 40%Scale 1-5

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

A highly actionable, well-sequenced orchestration skill with real reference files, explicit validation, and error-recovery paths. The main cost is token efficiency: duplicated step numbering, triple-stated confirmation rules, and inline contract detail push the body well past what its own reference files could absorb.

Suggestions

Fix the duplicated step numbering: merge "Step 0: Load Config" and "Step 0: Load Context" into a single Step 0 so the phase sequence reads unambiguously.

State the explicit-confirmation rule (criteria, max iterations, time budget) once as a guardrail and reference it from the quick-mode and full-setup lists instead of repeating it three times.

Compress the skill-context section to its core rule (skill-context overrides win; verify all outputs against them) and move the run.json/current.json schemas into a reference file, keeping only field-role summaries inline.

Dimension	Reasoning	Score
Conciseness	The body is dense and project-specific (no concepts Claude already knows), but contains real redundancy: two sections are both numbered "Step 0" ("Load Config" and "Load Context"), the explicit-confirmation rule is stated three times (quick-mode steps 5-8, the "Critical guardrail", and "Never treat criteria... as final"), and the skill-context section restates its override rule four different ways. It could be tightened without losing anything.	3 / 5
Actionability	Copy-paste-ready JSON templates for run.json and current.json, an exact history.jsonl example line, concrete run_id/alias formats, a mkdir command, a stop-reason-to-status mapping table, an exact iteration-summary output template, and invocation examples fully specify execution; deferring phase I/O details to the real PHASE-CONTRACTS.md is a well-signaled split, not a gap.	5 / 5
Workflow Clarity	Steps 0-9 are clearly sequenced with explicit validation checkpoints (EVALUATE scoring, retry-once with phase_error, corrupted-run.json reconstruction from history.jsonl), feedback loops (CRITIQUE->REFINE, stagnation tracking), and a precedence-ordered stop-condition contract with rationale. The duplicate "Step 0" numbering is a blemish but does not break the sequence.	5 / 5
Progressive Disclosure	Six reference files (ACTIVE-TIME-BUDGET, CONTEXT-MANAGEMENT, CRITERIA-TEMPLATES, PHASE-CONTRACTS, RULE-SCHEMA, TERMINAL-REPORT) all exist, are exactly one level deep, and are clearly signaled in bold at their point of use with an inline summary plus pointer. However, the ~500-line body still inlines full schemas and contract prose (run.json schema, stop-condition precedence, terminal-report rules) that could partly live in the existing reference files.	4 / 5
	Total	17 / 20 Passed

Description

83%Weight 40%Scale 1-5

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

A strong description with a clear what, an explicit and well-phrased "Use when..." trigger clause, and a mostly distinct niche. Weakest spots are the generic word "artifact" (what kinds of artifacts?) and slightly jargon-tinged trigger vocabulary.

Dimension	Reasoning	Score
Specificity	"Run a strict multi-iteration Reflex Loop with phases (PLAN, PRODUCE\|\|PREPARE, EVALUATE, CRITIQUE, REFINE)... until quality gates pass or iteration limits are reached" names several concrete actions (six phases, gating, iteration limits), but "improve an artifact" leaves the artifact actions generic, so it falls just short of comprehensive coverage.	4 / 5
Completeness	It explicitly answers both what ("Run a strict multi-iteration Reflex Loop with phases (...) to improve an artifact until quality gates pass or iteration limits are reached") and when ("Use when user asks for iterative refinement, quality-gated generation, or \"generate -> critique -> refine\" loops") with concrete trigger phrases, matching the top anchor; it is not a 4 because the when-clause is already explicit and specific.	5 / 5
Trigger Term Quality	"iterative refinement", "quality-gated generation", and "generate -> critique -> refine" loops give good natural keyword coverage, though common variations like "keep iterating", "self-critique", or "refine until it passes" are missing and "quality-gated generation" leans jargon.	4 / 5
Distinctiveness Conflict Risk	"Reflex Loop" plus the named phase pipeline is a distinct niche, but the trigger "iterative refinement" is broad enough to overlap with other refinement/critique skills, giving minor conflict risk rather than minimal.	4 / 5
	Total	17 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (507 lines); consider splitting into references/ and linking	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 2 missing	Warning

	Total	13 / 16 Passed

Repository: lee-to/ai-factory
Path: skills/aif-loop/SKILL.md
Commit: ac92beb

Reviewed: 4 days 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.