Files
bare-operating-system/kernel/share/agent-workspace/TOOLS.md
T
2026-08-18 15:51:13 -04:00

5.3 KiB

TOOLS.md - Available Capabilities

You already have these tools. Schemas are attached. NEVER ASK to use them.

Files / VFS

  • read_file — UTF-8, optional offset/limit numbered slices.
  • write_file — create or overwrite (parents created).
  • edit_file / search_replace — unique old_string unless replace_all.
  • apply_patch — Codex/Grok *** Begin Patch multi-file edits (Add / Delete / Update / Move).
  • create_directory, list_directory (tree=true for a bounded BFS tree), file_stat, glob_files.
  • grep — VFS-native regex search (glob, ignore_case, context, output_mode content | files_with_matches | count). Prefer over search_files / run_command grep.
  • find_symbol — definition-oriented search (function/class/const/def/fn).
  • search_filesgrep -Rnl via the guest shell (legacy).
  • move_path, copy_path, diff_files, delete_path — enabled. Base system is read-only.
  • list_bin — guest /bin utilities (POSIX-in-JS, not GNU).
  • git_statusgit status --short, diff --stat, recent log.
  • git_log / git_show / git_blame — prefer these over raw run_command git.

Prefer list_directory / glob_files / file_stat / grep over ls / find / shell grep.

Code / harness

  • run_command — any guest shell line (cwd optional; last cwd persisted). command_deny empty by default.
  • run_js_script — required for agent-authored JS. Node is not installed. Writes ~/.agent/_tmp_agent_run.mjs and runs it on the Bare kernel. Prefer async function run(ctx, argv).
  • run_js_script_at_path — existing absolute .mjs.
  • todo_write — session todos (merge=true to update by id).
  • enter_plan_mode / exit_plan_mode — plan mode is read-only except ~/.agent/plan.md.
  • memory_search / memory_get / memory_append / remember — MEMORY.md and daily logs.
  • list_skills / read_skill / create_skill — catalog, full SKILL.md, or author a new workspace skill.
  • Project skills are also walked from .grok/skills, .agents/skills, .claude/skills, .cursor/skills.
  • schedule_task / unschedule_task / list_scheduled — guest timers (agent-*.timer).
  • fuzzy_find — filename search when you remember part of a name.
  • read_many — several files in one call.
  • wait_for — poll a file or command until a regex matches (sync, timeout).
  • undo_last_edit — restore the last snapshot from ~/.agent/edits.json.
  • git_diffgit diff / --stat.
  • history_search — keyword search of this session's history.
  • rewind_session — Grok /rewind: drop the last N user turns from ~/.agent/history.json.
  • export_session — Grok /export: write a Markdown transcript (default ~/.agent/export.md).
  • glob/grep honor .gitignore, .agentignore, and .grokignore at the walk root.
  • edit_agent_config — shallow merge of known ~/.agent/config.json keys.

Kernel / ops

  • read_proc_file — any /proc path. Live features: /proc/bare_os/features or features.json, plus capabilities.json.
  • runtime_diagnostic_bundle — one-shot ctx + /proc/bare_os snapshot. Prefer this over many separate reads.
  • get_system_info, get_resource_limits, get_swarm_peers.
  • list_services, service_status, get_initd_graph, read_unit_journal.
  • list_timers, read_cron_log, read_audit_log.
  • read_boot_policy, read_kernel_extension_resolution.
  • inspect_ipc_backpressure, get_network_summary, tail_telemetry_streams, pkg_index_lookup.

Man pages (read_man_page, apropos_man) are documentation search only — never infer live kernel state from them.

Checks (guest wrappers + host hints)

  • list_verification_scripts, run_maintenance_gate, run_contract_checks, summarize_build_drift.
  • verification_hints — suggests host-checkout npm / node checks. Does not run them here.

Bridge / web

  • web_search — DuckDuckGo instant answers (same HTTP policy as wget/curl). Then web_fetch promising URLs.
  • web_fetch — http(s) via ctx.httpFetch (same host allow/deny as wget/curl).
  • get_hrpc_bridge_health, get_hrpc_allowlist_status.
  • emit_host_notification, request_host_action — enabled by default. emergency_stop_mutations can still block.

Autonomy

  • autonomous_run / autonomous_run_status / autonomous_run_stop / update_goal.
  • Guardrails are denylists only (autonomous_deny_ops empty by default). Keep the tool loop going.

Other

  • ask_user_question — product choice only, never permission.
  • task_complete — call when the user task is done.

Policy toggles (~/.agent/config.json)

  • access_policy: full (default) or restricted.
  • allow_delete, allow_bridge_mutations, allow_host_notifications, allow_host_actions: true.
  • command_deny, autonomous_deny_ops: empty by default.
  • mutate_deny_prefixes: read-only base-system prefixes.
  • emergency_stop_mutations: kill switch for mutating bridge tools.

Reasoning / process visibility

Configured in ~/.agent/config.json (show_reasoning, reasoning_mode, reasoning_max_chars, reasoning_include_tools). Progress UI is automatic from ~/.agent/progress.txt. Use edit_agent_config to toggle.

  • REST xai: https://api.x.ai/v1. Prefer trace mode when summaries are unavailable.
  • REST groq: https://api.groq.com/openai/v1; tool loops use parallel_tool_calls and max_completion_tokens.