/health-check

Validates that the AI ForgeKit environment is correctly structured and internally consistent. Reports PASS, WARN, or FAIL per check.

Trigger phrases#

/health-check, check environment health, is everything set up correctly, validate my environment, run diagnostics, or after updating the environment.

Do NOT use for: checking game build status (use /build) or reviewing code (use /review).

This skill is read-only#

/health-check validates and reports. It does not modify any files.

Two-layer validation#

Layer 1: Structural checks (18 checks, S1–S18)#

Run by a deterministic script that checks file existence, JSON validity, and naming conventions:

ID	Check	Severity
S1	CLAUDE.md exists in project root	FAIL
S2	.claude/settings.json parses as valid JSON	FAIL
S3	Each .claude/skills/*/ has a SKILL.md	FAIL
S4	Skill folder names are kebab-case	WARN
S5	No README.md inside skill folders	WARN
S6	Each .claude/agents/*.md has name, description, maxTurns	WARN
S7	Docs/Design/index.md exists	FAIL
S8	Docs/Dev/index.md exists	FAIL
S9	Docs/Dev/Systems/ directory exists	FAIL
S10	Docs/_working/state/ directory exists	WARN
S11	Docs/_working/retro/ directory exists	WARN
S12	CLAUDE.md is under 250 lines	WARN
S13	Each .claude/rules/*.md is under 120 lines	WARN
S14	Each .claude/agents/*.md is under 80 lines	WARN
S15	Each SKILL.md is under 500 lines	WARN
S16	settings.json has a non-empty deny array	WARN
S17	settings.json denies .uasset and .umap edits	FAIL
S18	project-config.json is valid JSON (if it exists)	WARN

Layer 2: Semantic checks (12 checks, L1–L12)#

Performed by reading and analyzing file content after the script completes:

ID	Check	Severity
L1	All file paths referenced in CLAUDE.md exist on disk	FAIL
L2	Routing table covers bugs, features, code questions, build issues, reviews, design queries, memory commands	WARN
L3	Routing entries use project-specific terms, not generic placeholders	WARN
L4	No contradictory rules across rule files	FAIL
L5	Docs/index.md retrieval guide matches actual files in Docs/Design/ and Docs/Dev/	WARN
L6	System and design pages with "Last Updated" older than 30 days	WARN
L7	Friction entries with 3+ occurrences and no corresponding proposal	WARN
L8	state-save covers all 6 taxonomy categories and state-load reads all of them	WARN
L9	settings.json allows all tools referenced by agents and skills	WARN
L10	No agents or skills referenced in routing table whose files do not exist	FAIL
L11	VCS rule file exists and lists dangerous commands	WARN
L12	Binary asset deny rules and .claudeignore patterns are present	WARN

Output format#

## Health Check Report

Run: YYYY-MM-DD HH:MM UTC
Environment: [Project name]

### Summary
PASS: N/N checks passed
WARN: N warnings
FAIL: N failures

### Failures (action required)
- [FAIL] S17: settings.json does not deny .uasset edits

### Warnings
- [WARN] L6: Docs/Dev/Systems/Combat/overview.md last updated 45 days ago

### Recommendation
[Fix FAILs first, then address WARNs. Suggest /update if retro patterns found.]

Detailed information on any check is available on request. Ask "what does S17 mean" or "details on L6" after the report.

When to run#

• After installing AI ForgeKit (verify structure is correct)
• After running /update (verify changes are consistent)
• After modifying any .claude/ rule or agent files
• When something seems wrong with agent behavior
• Periodically as a maintenance check

/health-check is the fastest way to confirm the environment is wired up correctly before starting a complex task.