Generate hierarchical AGENTS.md files. Root + complexity-scored subdirectories.
\n/init-deep # Update mode: modify existing + create new where warranted\n/init-deep --create-new # Read existing → remove all → regenerate from scratch\n/init-deep --max-depth=2 # Limit directory depth (default: 3)\nMark \"discovery\" as in_progress.
\nDon't wait-these run async while main session works. Equip every agent with the code graph: any task touching structure, entry points, dependencies, or hotspots MUST query codegraph_* (explore/search/callers/callees/impact) and lsp_symbols when present, and ground its claims in that data instead of guessing from conventions. Richer real-graph context per agent = a more accurate project map.
// Fire all at once, collect results later\ntask(subagent_type=\"explore\", load_skills=[], description=\"Explore project structure\", run_in_background=true, prompt=\"Project structure: map real layout via codegraph_explore/codegraph_files → REPORT deviations from standard patterns\")\ntask(subagent_type=\"explore\", load_skills=[], description=\"Find entry points\", run_in_background=true, prompt=\"Entry points: FIND main files, trace reach via codegraph_callees + lsp_symbols → REPORT non-standard organization\")\ntask(subagent_type=\"explore\", load_skills=[], description=\"Find conventions\", run_in_background=true, prompt=\"Conventions: FIND config files (.eslintrc, pyproject.toml, .editorconfig) → REPORT project-specific rules\")\ntask(subagent_type=\"explore\", load_skills=[], description=\"Find anti-patterns\", run_in_background=true, prompt=\"Anti-patterns: FIND 'DO NOT', 'NEVER', 'ALWAYS', 'DEPRECATED' comments → LIST forbidden patterns\")\ntask(subagent_type=\"explore\", load_skills=[], description=\"Explore build/CI\", run_in_background=true, prompt=\"Build/CI: FIND .github/workflows, Makefile → REPORT non-standard patterns\")\ntask(subagent_type=\"explore\", load_skills=[], description=\"Find test patterns\", run_in_background=true, prompt=\"Test patterns: FIND test configs/structure; codegraph_callers on core modules to see what is covered → REPORT unique conventions\")\nWhile background agents run, main session does:
\n# Directory depth + file counts\nfind . -type d -not -path '*/\\\\.*' -not -path '*/node_modules/*' -not -path '*/venv/*' -not -path '*/dist/*' -not -path '*/build/*' | awk -F/ '{print NF-1}' | sort -n | uniq -c\n\n# Files per directory (top 30)\nfind . -type f -not -path '*/\\\\.*' -not -path '*/node_modules/*' | sed 's|/[^/]*$||' | sort | uniq -c | sort -rn | head -30\n\n# Code concentration by extension\nfind . -type f \\\\( -name \"*.py\" -o -name \"*.ts\" -o -name \"*.tsx\" -o -name \"*.js\" -o -name \"*.go\" -o -name \"*.rs\" \\\\) -not -path '*/node_modules/*' | sed 's|/[^/]*$||' | sort | uniq -c | sort -rn | head -20\n\n# Existing AGENTS.md / CLAUDE.md\nfind . -type f \\\\( -name \"AGENTS.md\" -o -name \"CLAUDE.md\" \\\\) -not -path '*/node_modules/*' 2>/dev/null\nFor each existing file found:\n Read(filePath=file)\n Extract: key insights, conventions, anti-patterns\n Store in EXISTING_AGENTS map\nIf --create-new: Read all existing first (preserve context) → then delete all → regenerate.
Highest-signal source for the CODE MAP and the Symbol/Export/Reference scoring rows. Complementary, not alternatives - run BOTH when present, alongside the explore agents.
\nLSP - check lsp_status; model-facing names are lsp_status/lsp_symbols/lsp_find_references/lsp_goto_definition (some harnesses drop the lsp_ prefix):
lsp_symbols scope=\"document\" on each entry point -> file outline.lsp_symbols scope=\"workspace\", query by kind (class/interface/function) -> symbol inventory.lsp_find_references on top exports (line/character from the symbols result) -> reference centrality.codegraph - when codegraph_* tools exist (check codegraph_status); a first-class peer to LSP, NOT a last resort:
codegraph_explore -> overview; codegraph_callers/codegraph_callees/codegraph_impact -> centrality + blast radius for the scoring matrix; codegraph_search/codegraph_files -> symbol/file inventory.Only if NEITHER exists: explore agents + the ast-grep skill (sg), and mark centrality unmeasured in the CODE MAP.
// After main session analysis done, collect all task results\nfor each background task ID (`bg_...`): background_output(task_id=\"bg_...\")\nMerge: bash + LSP/codegraph + existing + explore findings. Mark \"discovery\" as completed.
\nMark \"scoring\" as in_progress.
\n| Factor | \nWeight | \nHigh Threshold | \nSource | \n
|---|---|---|---|
| File count | \n3x | \n>20 | \nbash | \n
| Subdir count | \n2x | \n>5 | \nbash | \n
| Code ratio | \n2x | \n>70% | \nbash | \n
| Unique patterns | \n1x | \nHas own config | \nexplore | \n
| Module boundary | \n2x | \nHas index.ts/init.py | \nbash | \n
| Symbol density | \n2x | \n>30 symbols | \nLSP/cg | \n
| Export count | \n2x | \n>10 exports | \nLSP/cg | \n
| Reference centrality | \n3x | \n>20 refs | \nLSP/cg | \n
| Score | \nAction | \n
|---|---|
| Root (.) | \nALWAYS create | \n
| >15 | \nCreate AGENTS.md | \n
| 8-15 | \nCreate if distinct domain | \n
| <8 | \nSkip (parent covers) | \n
AGENTS_LOCATIONS = [\n { path: \".\", type: \"root\" },\n { path: \"src/hooks\", score: 18, reason: \"high complexity\" },\n { path: \"src/api\", score: 12, reason: \"distinct domain\" }\n]\nMark \"scoring\" as completed.
\nMark \"generate\" as in_progress.
\n\n# PROJECT KNOWLEDGE BASE\n\n**Generated:** {TIMESTAMP}\n**Commit:** {SHORT_SHA}\n**Branch:** {BRANCH}\n\n## OVERVIEW\n{1-2 sentences: what + core stack}\n\n## STRUCTURE\n/\n├── / # └──
\n\n## WHERE TO LOOK\n| Task | Location | Notes |\n|------|----------|-------|\n\n## CODE MAP\n{From LSP/codegraph - skip only if neither exists or project <10 files}\n\n| Symbol | Type | Location | Refs | Role |\n|--------|------|----------|------|------|\n\n## CONVENTIONS\n{ONLY deviations from standard}\n\n## ANTI-PATTERNS (THIS PROJECT)\n{Explicitly forbidden here}\n\n## UNIQUE STYLES\n{Project-specific}\n\n## COMMANDS\n```bash\n{dev/test/build}\n\n**Quality gates**: 50-150 lines, no generic advice, no obvious info.\n\n### Subdirectory AGENTS.md (Parallel)\n\nLaunch writing tasks for each location:\n\nfor loc in AGENTS_LOCATIONS (except root):\ntask(category=\"writing\", load_skills=[], run_in_background=false, description=\"Generate AGENTS.md\", prompt=Generate AGENTS.md for: ${loc.path} - Reason: ${loc.reason} - 30-80 lines max - NEVER repeat parent content - Sections: OVERVIEW (1 line), STRUCTURE (if >5 subdirs), WHERE TO LOOK, CONVENTIONS (if different), ANTI-PATTERNS)
\n**Wait for all. Mark \"generate\" as completed.**\n\n---\n\n## Phase 4: Review & Deduplicate\n\n**Mark \"review\" as in_progress.**\n\nFor each generated file:\n- Remove generic advice\n- Remove parent duplicates\n- Trim to size limits\n- Verify telegraphic style\n\n**Mark \"review\" as completed.**\n\n---\n\n## Final Report\n\n=== init-deep Complete ===
\nMode: {update | create-new}
\nFiles:\n[OK] ./AGENTS.md (root, lines)\n[OK] ./src/hooks/AGENTS.md ( lines)
\nDirs Analyzed: AGENTS.md Created: AGENTS.md Updated:
\nHierarchy:\n./AGENTS.md\n└── src/hooks/AGENTS.md
\n\n---\n\n## Anti-Patterns\n\n- **Static agent count**: MUST vary agents based on project size/depth\n- **Sequential execution**: MUST parallel (explore + LSP + codegraph concurrent)\n- **Ignoring existing**: ALWAYS read existing first, even with --create-new\n- **Over-documenting**: Not every dir needs AGENTS.md\n- **Redundancy**: Child never repeats parent\n- **Generic content**: Remove anything that applies to ALL projects\n- **Verbose style**: Telegraphic or die\n