agent-harness-evolve

Evoluciona un arnes con lo aprendido: extrae reglas desde feedback de revision hacia AGENTS.md, captura playbooks reutilizables sin PII, y aplica condiciones de retiro a los documentos de gobierno cumplidos (trabajo terminado en pasado, pendiente en imperativo). Usar al pedir 'evolucionar arnes', 'extraer reglas', 'playbook del proyecto', 'retirar plan cumplido'. NO para fixes mecanicos (agent-harness-repair).

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

82%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 terse, well-structured procedural skill that assumes harness literacy and includes real validation checkpoints. Its main weaknesses are an abstract rule-authoring step, a missing error-recovery loop after re-audit, and a Packet block listing bundle layers that don't exist here.

Suggestions

Add an explicit recovery branch after step 5: if Q12 stays red (retiro not declared), state the remediation rather than only the green-pass condition.

Make the rule-authoring step more copy-paste concrete by showing the expected AGENTS.md structured-line format inline or via a short example in references/reglas-y-retiro.md.

Reconcile the Packet layer list with the actual bundle: either remove knowledge/, prompts/, examples/, agents/ from the wire-packet block or note they are optional/external so the listing reflects what is shipped.

Dimension	Reasoning	Score
Conciseness	Lean, jargon-dense sections with no padding; assumes Claude knows Q9/Q12, ExperimentSpec v2, archive_location, and AGENTS.md structure without explaining them — every token earns its place.	5 / 5
Actionability	Concrete procedural steps naming specific mechanisms (AGENTS.md structured line, ExperimentSpec v2 validate_spec, archive_location); as an instruction-only skill no-code is acceptable, but the rule-authoring step is somewhat abstract rather than copy-paste ready, keeping it below 5.	4 / 5
Workflow Clarity	Six sequenced steps with a validation checkpoint (step 5 'Re-audita: Q12 en verde') and an adoption gate (step 6 bounded experiment); the operation is explicitly non-destructive (archives vs deletes) so the destructive cap does not apply, but no explicit error-recovery loop is given if Q12 stays red.	4 / 5
Progressive Disclosure	Clean sectioned overview with a well-signaled one-level reference to the real references/reglas-y-retiro.md; the Packet block advertises layers (knowledge/, prompts/, examples/, agents/) that are not present in this bundle, a minor organization gap below the top anchor.	4 / 5
	Total	17 / 20 Passed

Description

87%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, specific description that names concrete actions, supplies natural trigger phrases, and explicitly disambiguates itself from the related repair skill. Minor keyword-synonym coverage is the only thing keeping it from a perfect specificity/trigger score.

Dimension	Reasoning	Score
Specificity	Lists three concrete actions with named mechanisms (extract rules into AGENTS.md, capture playbooks without PII, apply retirement conditions to fulfilled governance docs); not a 5 because coverage is solid but not exhaustive of every sub-operation.	4 / 5
Completeness	Explicitly answers both 'what' (the three actions) and 'when' (Usar al pedir...) with concrete trigger phrases and a negative boundary; matches the top anchor.	5 / 5
Trigger Term Quality	Provides several natural trigger phrases a user would say ('evolucionar arnes', 'extraer reglas', 'playbook del proyecto', 'retirar plan cumplido'); a few common synonyms/variations are absent so it stops short of 5.	4 / 5
Distinctiveness Conflict Risk	Clear niche (harness evolution) with an explicit boundary clause ('NO para fixes mecanicos (agent-harness-repair)') minimizing conflict with the sibling repair skill.	5 / 5
	Total	18 / 20 Passed

Validation

68%

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 — 11 / 16 Passed

Validation for skill structure

Criteria	Description	Result
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning
metadata_version	'metadata.version' is missing	Warning
metadata_field	'metadata' should map string keys to string values	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	11 / 16 Passed

Repository: JaviMontano/claude-plugins
Path: plugins/claude-native-toolkit/skills/agent-harness-evolve/SKILL.md
Commit: e8f986b

Reviewed: 3 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.