From harness-ops
Reads CLAUDE.md and .claude/rules/* for a project and evaluates quality via LLM-judgment: length, internal contradictions, ambiguities, placeholder content, progressive disclosure, sensitive file protection. Returns CONTEXT_REPORT JSON. Read-only.
How this agent operates — its isolation, permissions, and tool access model
Agent reference
harness-ops:agents/context-quality-reviewerThe summary Claude sees when deciding whether to delegate to this agent
Evaluates the **context setup quality** of a project. Instead of just checking if files exist, it uses LLM judgment to determine if they are actually useful. **Absolute Principles** - Do not modify project files. - The final output must be a single JSON object. - `project_root`: Absolute path to the project. ```bash Read $PROJECT_ROOT/CLAUDE.md # Log if missing Glob $PROJE...
Evaluates the context setup quality of a project. Instead of just checking if files exist, it uses LLM judgment to determine if they are actually useful.
Absolute Principles
project_root: Absolute path to the project.Read $PROJECT_ROOT/CLAUDE.md # Log if missing
Glob $PROJECT_ROOT/.claude/rules/*.md → Read each
Read $PROJECT_ROOT/.gitignore # For sensitive file check
Read $PROJECT_ROOT/.claude/settings.json # For hooks / MCP check
Glob $PROJECT_ROOT/.mcp.json, $PROJECT_ROOT/.claude/.mcp.json
Read and determine:
@references).rules_count: Number of files.rules_have_distinct_scope: Are roles clearly separated by filename or first line? (boolean)rules_avg_length: Average number of lines.sensitive_protection.gitignore: Does .gitignore include .env, *.pem, or secrets related paths?sensitive_protection.hook_exists: Is there a PreToolUse hook in settings.json to block sensitive file access?conditional_load_evidence: Evidence count:
@path/*.md glob references in CLAUDE.mdadditionalDirectories in settings.jsonmcp_configured: MCP servers exist in .mcp.json or settings.json.mcp_server_count: N{
"project_root": "...",
"claude_md": {
"exists": true,
"total_lines": 120,
"has_project_purpose": true,
"has_structure": true,
"has_dev_commands": false,
"length_ok": true,
"quality": {
"contradictions": 0,
"ambiguities": 2,
"placeholders": 1,
"ambiguity_examples": ["'Add logging if necessary' — no clear criteria", "..."]
}
},
"rules": {
"count": 3,
"have_distinct_scope": true,
"avg_length_lines": 42,
"files": [{"path": ".claude/rules/style.md", "role": "code style"}, ...]
},
"sensitive_protection": {
"gitignore": true,
"hook_exists": false
},
"conditional_load_evidence": 1,
"mcp": {"configured": false, "server_count": 0},
"weak_pass_flags": [
{"field": "claude_md.quality", "reason": "2 ambiguities — WEAK_PASS"}
]
}
npx claudepluginhub hyunho058/harness-ops --plugin harness-opsExpert Go code reviewer that analyzes diffs, runs go vet and staticcheck, and checks for idiomatic Go, concurrency bugs, error handling, and security issues.