NLPM Audit: 5 documentation bugs found in skills/ (score 89/100)

Automated audit: This issue was generated by NLPM, a natural language programming linter, running via claude-code-action. Please evaluate the findings on their merits.

Overview

KubeSphere's skill files scored 89/100 on NLPM's 100-point natural language programming quality scale. That's a strong result — the skills are well-structured with clear prerequisites, good scoping, and exemplary patterns in several files.

NLPM's scanner found 5 mechanical bugs (not style issues) that can cause agent misbehavior or invalid output. Each has a dedicated PR with a minimal, targeted fix.

Methodology

NLPM applies a deterministic 100-point penalty table to NL artifacts (skills, agents, prompts). Bugs are findings where the artifact's own stated rules contradict its examples, or where a syntax error would produce invalid output. Only items in the "Bugs" category are PR-worthy; style and quality issues are informational only.

Bugs Found (PRs submitted)

#	File	Bug	PR
1	skills/kubesphere-devops-credentials/SKILL.md	API examples use "type": "Opaque" — contradicts the file's own warning that Opaque breaks Jenkins credential sync (controller ignores Opaque secrets entirely)	#6632
2	skills/kubesphere-fluid/SKILL.md	YAML template has unclosed double-quote: low: "{{low}} → missing closing " produces invalid YAML	#6633
3	skills/whizard-logging/SKILL.md	Placeholder comment says From Step 3 but only Step 1 and Step 2 are defined; agents look for a non-existent step	#6634
4	skills/kubesphere-devops-pipeline/SKILL.md	**Step 3b: For Private Repository (with credential): is missing closing **; unclosed bold renders incorrectly	#6635
5	skills/kubesphere-devops-overview/SKILL.md	Project Components ASCII architecture diagram duplicated verbatim; second copy appears immediately before ## Installation	#6636

Priority

Bug #1 (credential type) is the highest-impact: agents following the examples will create credentials that silently fail to sync with Jenkins, causing pipeline failures with CredentialId could not be found. The remaining bugs are lower-impact formatting and reference issues.

What was NOT submitted

• Quality/style issues (duplicate content in argocd, vague quantifiers) — informational only, no PRs
• High-severity security finding (xargs shell injection in config/ks-core/charts/ks-crds/scripts/post-delete.sh) — practical risk is low given Kubernetes resource name constraints, but flagged here for awareness; please review at your discretion

Thank you for maintaining such a well-documented skill library. Happy to clarify any findings or close this issue if the PRs are not welcome.